'(!I",., "

tit:'~-'"" ~,. .

'. ~.

.~: .. bI

.....\~..,.

SENTENCE PROCESSING: Psycholinguistic Studies Presented to Merrill Garrett

EDITED BY

WILLIAM E. COOPER Harvard University

EDWARD C. T. WALKER Massachusetts Institute of Technology

m LAWRENCE ERLBAUM ASSOCIATES, PUBLlSHH' 1979 Hillsdale, New Jersey

DISTRIBUTED BY THE-HALSTED PRESS DIVISiON OF ,

JOHN WILEY & SONS'~

342 SHATTUCK-HUFNAGEL

Shattuck-Hufnagel, S., & Klatt, D. H. Phonetic speech errors rule our three model" (,r'l'ml'llC<'

production. Paper prepared for the working group on slips of the tongue (V. Fromkin, Chairperson). XII International Congress of Linguists. Vienna. Austria, September 1977.

Shattuck-Hufnagel, S., & Klatt. D. H. The limited use of distinctive features and markedness in speech production: Evidence from speech error data. Journal of Verbal Learning and Verbal Behavior, 1978, in press.

Shattuck-Hufnagel, S.• & Klatt. D. H. Feature constraints on different types ofphoneticspeech errors. (Manuscript in preparation).

Wickelgren, W, A. Distinctive features and errors in short-term memory for English consonants. Journal of the Acoustical Society of America. 1966.39,388-398.

Like Syntax

John Robert Ross Massachusetts Institute of Technology

William E. Cooper Harvard University

In developing a theory of ongoing speech production, researchers have typically assumed that important clues may be provided by studying factors that occur in casual speech but not in writing. This assumption serves as a basis for investigating speech errors (e.g., Fromkin, 1971; Garrett, 1975; Shattuck-Hufnagel, Chapter 10 of this volume), hesitation pauses (e.g., Boomer, 1965;Goldman-Eisler, 1968;Martin, 1971),and various phonologi­cal reductions (Baker, 1972; King. 1970; Starnpe, 1972; Zwicky. 1970). Here. we invoke the assumption as a rationale for studying restrictions on the occurrence of the word like as a filler. meaning roughly for example or approximately, as in (I).

(I a) John is like the best baseball player in town. (I b) We intended to leave Chicago at like 9 o'clock.

The precise meaning of like as a filler varies in different sentences (a matter we will return to later). Our primary concern, however. lies not with these shades of meaning but with restrictions on like's occurrence.

In many sentences. including (I )./ike can occur in casual speech without all accompanying pause. In this respect, like's behavior can be distinguished from interjections (c.g., oh, \\,('I!. say) studied by James (197)), In other contexts. as in (2). a bordering pause is required:

(2a) Bob's like *(,) father would not leave town without us. (2b) We should leave town *(,) like,,~~..•~ ,1JIiI

~'~_~~~~W~~~~~~.¥~~~,: :~ ~.

343

• •

344 ROSS AND COOPER 11. LIKE SYNTAX 345

In (2) and throughout this paper, we denote a pause by a comma. An asterisk is placed outside the parenthesized comma to signify that the sentence is ungrammatical when the comma is not present.

Initial1y, we confine our remarks to like as it occurs in contexts that do not require a pause. We then consider pause-bounded like and relate like's behavior to adverbial intensifiers such as even, only, also, and just. The analysis of like and intensifiers provides consequences for the analysis of English auxiliaries and other issues. Finally, we discuss implications for a theory of sentence processing.

PAUSELESS LIKE

Let us examine a typical example of the kind of puzzle that like poses to any researcher. Consider (3):

(3) >..Why>..don't'>.,we>..call>..both>..Harriet>..and>..Suzie>..up

7·· •• . 7· • >.. this>..afternoony7

The results of inserting one or more likes in the various niches between the words of (3) are given by what appears under the lambdas in (3): if nothing is there, we would find grammatical a sentence appearing with like in the position of the lambda. On the other hand, if an asterisk or some other index of less than perfect grammaticality is under a lambda, like in that place will produce a sentence that is ungrammatical to that extent.

A number of factors can be isolated that predict such arrays of asterisks as appear in (2) and (3). Let us briefly examine some.

First, as Bill Cantrall has pointed out to us (personal communication. 1977), because like often means something close to "approximately," it will be semantically odd if it precedes any element about which the speaker must be certain [examples due to Cantral1J:

(4a) Give 'em hell, (·Iike) Harry! (Speakers know to whom they are speaking.)

(4b) What's (·Iike) that? (out when like is not a preposition) (4c) May I introduce you to (·like) my mother? (out as a bona fide

introduction)

'When the pronoun receives emphatic stress. however, like is permitted. This observation i' accounted for in thc following discussion by reference to like's association with focus.

Although these examples argue strongly that one component in a description of like must be a semantic one, there are equally strong arguments to the effect that phonology must play 'a role.?This appears most clearly in cases involving two synonymous expressions, one of which is longer than the other, such as submarine and sub, or full names and nicknames. .

(5a) They photographed two like submarines. (5b) ?They photographed two like subs. (6a) Did they invite like Isadore? (6b) ?·Did they invite like IZ?

Of course, the short and long expressions also differ in their degree of formality, but like's preference for casual speech would predict, if anything, that (5b) and (6b) would be better than (5a) and (6a), contrary to fact. It seems probable to us that the reason for these contrasts is phonetically determined; i.e., after like. an elaborate fundamental frequency gesture must be executed and that with a short closed syllable, there is not enough time to perform this gesture.

I I The occurrence of like is also influenced by overall speaking rate, with

greater acceptability at faster rates. Thus, like can only appear in sentences spoken at a fairly rapid rate of speech, e.g., (7a), and not in the same sentence read with exaggerated slowness, as in (7b).

(7a) I was (like) very disappointed in you. (7b) I ... was ... (·Iike) ... very ... disappointed ... in ... you.

It appears that like is often used in fast speech because of the speaker's need to buy processing time.

To demonstrate that a full description of the behavior of /ike will have to be sensitive to both semantic and phonological parameters, we examine cases of

lDennis Peacock (personal communication. 1977)has suggested that the preferred position of like is at the first +juncture to the left of a stress increase. Such an account would correctly block like before the head noun of an NP. In addition. according to Peacock. the juncture account would block like from occurring between two words not separated by a -juncture, as in (ib) below, where want 10 is not separated by a -juncture and is reducible to «anna.

(ia) He wants like to do it alone. (Peacock's judgments) (ib) '. want like to do it alone.

In our speech, however, (ia) and (ib) are equally acceptable. Also, the -juncture account fails to predict the acceptability of cases such as the following. where like can be inserted at a location where no +juncture exists, as evidenced by the reductions possible when like is omitted:

(iia) I got like your message. (KOI .l'0llr-Kolcha)

(iib) Let's aid like your mother. (aid your d-J palatalization)

"'0:Fi1iJWtIi-~ ."--,;r,.,r­ . •• ie lIIsl _1JiI.·1Q1!~ ...

:1i.'<...,- ..._.~.

/ 346 JS AND COOPER

constituent structure ambiguity. Consider (Sa), in which the prepositional phrase 10 Kennedy either can be the indirect object of send, on which reading (Sa) and (8b) are synonymous, or can be an object of the adjective insulting, on which reading it is (roughly) synonymous with (8c).

(8a) The pranksters sent the letters that were insulting to Kennedy. (8b) The pranksters sent Kennedy the letters that were insulting. (8c) The letters that were insulting to Kennedy were sent by the

pranksters.

Note that like is often associated with elements that bear emphatic stress, as Kennedy would, in (8a), if it were used to answer (9).

(9) To whom did the pranksters send the letters that were insulting?

What is of interest here is that, although like could precede to Kennedy in an answer to (9), it could not precede insulting: thus, (lOa) can answer (9), but (lOb) cannot.

(J Oa) The pranksters sent the letters that were insulting like to Kennedy. (lOb) '"The pranksters sent the letters that were like insulting to Kennedy.

On the other hand, there are contexts in which (lOb) could be used; e.g., as an answer to (J I).

(II) What kind of letters did the pranksters send?

I n an answer to (II). insulting to Kennedy would be a constituent, whereas it would not be one in an answer to (9). This type of syntactic difference is the kind that we, rather arbitrarily. single out as a starting point for our examination of the behavior of like.

Pauseless like is typically prohibited before certain grammatical categories. including head nouns. tensed auxiliaries. and pronouns, as in (12):

( 12a) "Hob 's like bank is closed on Fridays. ( l2b) *We saw a rcd like helicopter at thc state fair. ( 12c) "Ed like is working nights. ( 12d) *It like has been raining for two days. ( 12e) *Wc gave like her a birthday party. ( 12f) *Did like you have a good round?

However, a momcm 's reflection serves to indicate that dcscribing construims in terms of grammatical categories per se is the wrong tree to bark up. Each of

\ 11. LIKE SYNTAX 34i

the sentences in (12) becomes accepta ble when the word following like is emphatically stressed, as in (13):

(l3a) Bob's like BANK is closed on Fridays. (13b) We saw a red like HELICOPTER at the state fair. (l3c) Ed like IS working nights. (l3d) It like HAS been raining for two days. (l3e) We gave like HER a birthday party. (131) Did like YOU have a good round?

Thus, we seem to be driven to the conclusion in (14):

(14) Pauseless like precedes and is associated with focused elements.

By focus, we mean new information contained in an utterance that is not assumed by the speaker to be previously shared by both speaker and hearer (following Chomsky, 1971;Jackendoff, 1972). Although it is not always clear exactly which element in a sentence contains focus. it is generally agreed that, in question-answer cases, the focus is associated with that part of the answer that corresponds to the wh word of the question, as in (15) through (17).

(l5a) Who is sending eggs to Marie? Focus

(15b) They are sending eggs to Marie. Focus

(l6a) What are they sending to Marie? Focus

(16b) They are sending eggs to Marie. Focus

(17a) Who are they sending eggs to? Focus

(17b) They are sending eggs to Marie. Focus

As Chomsky (1971) has noted, the focused clement need not correspond to a constituent in deep structure. Thus. like can be associated with the focused phrase "certain to lose" in the following sentence in reply to the question "Is John certain to win?"

(18) No, John is like certain to lose.

In this case, the approximate deep structure is:

(19) [sJohn win], is certain. = Chomsky's [57]

\,i;;f.:~'Ii~'iB~'1I.r.':9~~~~I""I1S~:·\~~"';:';!;'~W~Y{~f"',r;'!"!"l~!.~·"· "'t~t:", -, ,.

·.:~~t.fi~~~t~j::~ ~,

348 ROSS AND COOPER

In the a sentences of (15) through (17), it is clear that the wh-words can be t preceded by like:

IL ~~

(1 (20a) Like who is sending eggs to Marie?

.)~

~f

.ft

~.Ji (20b) Like what are they sending to Marie? (20c) Like who are they sending eggs to?

The claim made by (14) is that in the b sentences like can only modify the focused elements.That it can modify these elements can be seen in (21); that it cannot modify other elements can be,seen from the ungrammaticality of the sentences in (22), when these are understood as answers to (/5a) through (17a).

(2Ia) Like THEY are sending eggs to Marie. (21 b) They are sending like EGGS to Marie. (2Ic) They are sending eggs to like MARIE. (22a) %THEY are like sending eggs to Marie. ' (22b) *THEY are sending like eggs to Marie. (22c) *THEY are sending eggs like to Marie. (22d) *THEY are sending eggs to like Marie.

We have prefixed (22a) with the percentage sign to indicate our belief that there is a good deal of variation in judgments of acceptability for this type of sentence. Whereas (22a) is dubious for us, there are speakers who accept it. In addition, there are some sentences that do not differ from (22a) in any significant respect, as far as we can see, in which a like following the focused element is acceptable for us. One such example is (23b).

Sli'it. ~; (23a) What sort of thing might appear, if we say this spell?

i1A (23b) A DEMON might like appear.

As far as we know, pauseless like can follow the focused element 'in this specific context only: The focus is on the subject noun phrase (NP), and like

,.'" follows the tensed auxiliary. When like is further away from the subject. as in t,i

~: (22b) through (22d), we know of no speakers who accept the sentences. Let us now turn to a consideration of the positions in which like can appear r. in sentences such as (16) and (17), where the focused N P is not a subject. The

,'. relevant facts appear in (24) and (25), respectively.

(24a) Like they are sending EGGS to Marie. (24b) *They like are sending EGGS to Marie. (24c) They are like sending EGGS to Marie. (24d) They are sending like EGGS to Marie. (24e) *They are sending EGGS like to Marie.

11, LIKE SYNTAX 349

(24f) *They are sending EGGS to like Marie. (24g) *They are sending EGGS to Marie like. (25a) Like they are sending eggs to MARIE. (25b) *They like are sending eggs to MARIE. (25c) They are like sending eggs to MARIE. (25d) *They are sending like eggs to MARIE. (25e) They are sending eggs like to MARIE. (25f) They are sending eggs to like MARIE'. (25g) *They are sending eggs to MARIE like.

Initially, this distribution of asterisks is puzzling, but we believe the basic principles are fairly simple. First, the ungrammaticality of(24e), (24f), (24g), and (25g) is due to the subpart of principle (/4) which states that like must precede the focus. We formulate this law as in (26).

(26) "Like" First (LF) Pauseless like must precede the element it modifies (except that for some speakers, like may follow a tensed auxiliary, yet be modifying a focused subject).

However, if like could appear anywhere to the left of the focused element, the sentences (24b), (25b), and (25d) would all be well formed. We believe that the reason for their ill-formation is a stronger law, given in (27).

(27) The "Like" as a Left-Bracket Condition (LLBC) Pauseless like must open a constituent that dominates the focused element.

Condition (27) wil1explain the ungrammaticality of(24b) and (25b), under the assumption that sentences with auxiliaries are ternary branching, as suggested in (28), which is presented in Fig. 11.1.

It is immaterial to our analysis whether there is a node auxiliary (Au x) that is distinct from verb (V), whether there is a node verb phrase (VP) or not. A II that is required for the LLBC to explain why like can follow but not precede' are is a structure that assigns constituency to the string of words from sending to MARIE and does not assign such status to the string of words from arc' to MARlE. Because this last node, NP J , is the focused element (we have circled it for dramatic effect), the LLBC will allow like to open NP.1 itself. or Sz or SI. Because V I is not one of these, like cannot precede it, cf. *(24b) and *(25b). and the same is true of NPi, which explains the badness of *(25d).

A fairly clear argument for the correctness of Condition (27) is provided by the structures of (29a) and (29b):

(29a) red and white and blue (29b) red, white, and blue.

1,,=-1,

/' -, \

11, LIKE SYNTAX 351

We assume that (29a) can be assigncd the coordinate, ternary-branching structure shown in (30), which is demonstrated in Fig, 11.2,as wcll as the two

w binary-branching structures shown in (J I), presented in Figs, II.J and 11.4­This claim (if", three-way structural ambiguity for(29a) is uncontrovcrsial. as far as we know, as is the claim, embodied in (32), that only the first of these structures, (30), can lose its first and by the operation of Conjunction

o, /®(~ Deletion,

~<--...... Q.._- 0 Adj ­""""

­ /~'----~," , .. ~ "

Adi, /. /.

~

c, > _a:'/~ TAT TArz " 0" <. -;- ,OJ' ",~ FIG, 11,2. Example (30) red and white and blue

V1 ',, !. , 00,, !::!. ,, 0:

Q) Adj , '" Q.

>'---~ e,, .. , " ><" tlJ, .>.,, Adjl Ad,,­'"

~

.~ ~

~

o u:::

-;;<~uil--\,-i ~

\ >- Q,l

, '" \ :.= \ :!.. ,, , \ ­,~ ,.,

,z -- .0: " ,

/~ Conj Adj

~" Adj2 /\\

T' At , .... FIG.11.3. Example(3Ia) red and while and blue

Adj /-, , -:

IIdl Adl

/',.; .­-'

1 Adl, Conj Adl2 Coni AdJ 3

I I I I I FIG.11.4. Example (J I h) red and white and blue

350

:"'l~~~~~:r;r'". .~&il~~ta~;1"l

352 ROSS AND COOPER

(32) Conjunct ion Delet iO/1 When 11 elements of a coordinate structure, n ~ 3, are separated by n - I occurrences of coordinating conjunctions, such as and or or, the first n - 2 of these may be deleted and replaced by a (sometimes optional?) "comma intonation."

Supposing, now, we focus on blue in (29), by forming the echo questions shown in (33).

(33a) red and white and WHAT? (33b) red white and WHAT?

We assume that although the first of these -is still three-ways structurally ambiguous, the second has only the coordinate structure corresponding to that in (30). Under this assumption, the differential behavior of /ike in (33) can be explained: cf. (34).

(34a) (like) red and (like) white and (like) WHA11 (34b) (like) red, (*like) white and (like) WHA11

The LLBC will only allow a /ike before white if white and WHAT is a constituent, as is the case in a structure such as (31a). However, in order for Conjunction Deletion to be able to remove the first and of (33), its structure must have been like (30), and in this structure, a like before white could not open a constituent that dominates WHA T. Thus, it appears that the LLBC properly characterizes a major constraint on like's occurrence.

PAUSE-BOUNDED LIKE

In summary, pauseless like precedes the focused element and opens the constituent containing this element. Let us turn now to like as it occurs with an accompanying pause. Recall that in sentences (24) and (25), the focused elements could be immediately preceded, but not followed, by pauseless like, In sentence (35), however, we see that the same focused elements can be followed, but not preceded, by pause-bounded like:

(35a) THEY, like, are sending eggs to Marie. (35b) They are sending EGGS, like, to Marie. (35c) They are sending eggs to MARl E, like.

In general, whenever pauseless like can immediately precede the focused element, we can find a synonymous grammatical sentence in which pause­

11, LIKE SYNTAX 353

bounded like immediately follows the focused element. To account for this highly systematic distribution, we propose the following transformational rule:

(36) "Like" Hopping X like A y

Structural OPTIONAL Description 2 3 4 Structural Change I 0 3+, +2+,4 Condition: A is a constituent that contains the focused element (as

determined by the LLBC)

For a clear demonstration that hopped like must follow the focused element, reconsider sentence (25d), reproduced below:

(25d) They are sending like eggs to MARIE.

Now note that if like is hopped around eggs, the sentence is still ungrammatical: cf. (3'7).

(37) *They are sending eggs, like, to MARIE.

Contrast (37) with (25e), which we have repeated for convenience as (38).

(38) They are sending eggs like to MARIE.

Despite the fact that (37) and (38) contain the same sequence of words, we see that they differ in grammaticality because, although the pauseless like of (38) can only be associated with a following focused element, the pause­bounded like of (37) can only be associated with a preceding focused element, and in (37) MARIE follows pause-bounded like.

The same point can be made another way: In (39b), we give the sentence that would result if we applied "Like't-Hopping to (24d), which we repeat as (39a).

(39a) They are sending like EGGS to Marie. (39b) They are sending EGGS, like, to Marie.

In *(40), we have repeated the ungrammatical *(24e), which is identical to (39b), except that there are no commas bounding like.

(40) "They are sending EGGS like to Marie.

~-

~f

5

354 . -'SS AND COOPER

The generalization that emerges from these two pairs of string-identical sentences (actually, all four are identical if we disregard intonational factors) is stated in (41).

(41) Pauseless like must precede the focused element it isassociated with: pause-bounded like must follow it.

Condition (41) predicts that like must be pauseless when it appears in sentence-initial position, whereas it must be bounded by a pause in sentence­final position. The prediction for sentence-final position isconfirmed by(42).

(42) George and I are leaving tomorrow" (,) like. . ,. , ~

the prediction for sentence-initial position fails, however, as shown in (43).I;! i :! (43) Like (,) George and I are leaving tomorrow. ~i '.'I We suspect that the pause allowable in (43) is a hesitation pause,

distinguishable from the pause normally accompanying like when it is hopped, although we cannot at present provide any support for such a distinction. We believe that the pause in (43) would be empirically distinguishable by its longer duration, compared with the duration of pauses that accompany hopped like.

Thus, it appears that, in general, pause-bounded like must directly follow the constituent that contains the focused element: This is the only kind of situation that our hopping rule (36) allows. Thus, (36) will account for the badness of "'(44d).

(44a) Who is working on Greek? (44b) Like BOB is working on Greek. (44c) BOB, like, is working on Greek. (from (40b], via "Like"lIo!Jpillgj (44d) "'BOB is, like, working on Greek.

Note, however. that (44a) can be answered by (45).

(45) BOB is working on Greek, like.

If like 80B is a constituent, an NP, in (44b), "Like" Hopping will be unable to apply to produce (45). Suppose, however, that like is not attached to BOB. but merely opens the sentence, as shown in (46), shown in Fig. 11.5.

The l.LBC, (27), would allow like to he associated with !lOR, in this «I"', yet "Like" l!o!J!Jing would still he unable to produce (45). The only \\a\ !<'

generate (45) hy (36), while preserving the LLBC, would he to postulate that

FIG. 11.5. Example (46)

FIG.11.6. Example (47)

11. LIKE SYNTAX 35

/ ~~/1'",,// I "-.//

like NP IS '" .>. II L ..... _..._._:.,

BOB working on Greek

5,

lik~ ~ BOB is working on Greek

"BOB is working on Greek" can be a constituent in (44b), presumably in some structure such as (47), shown in Fig. 11.6.

The conclusion that an immediately self-dominating structure is necessary to allow "Like" Hopping to operate correctly can be supported by simpler cases of hopped like. Consider (48b), the hopped version of (48a), which is an alternate answer to (44a).

(48a) Like THE LINEBACKER is working on Greek. (48b) The LINEBACKER, like, is working on Greek.

Again, if the formulation in (36) is correct, the subject of (48a) must have a structure such as that shown in (49b); (49a) would not provide a constituent for like to hop over (see Fig. J1.7).

Let us return now to the prediction implicit in our formulation of "Like" Hopping: That a pause-bounded like must immediately follow a focus­containing constituent. This prediction correctly excluded "'(44d), and, similarly, it can explain why (SOc) cannot be used as a like-hopped version of (SOb), in replying to (50a).

NP NP,

.>: Aplike THE LINEBACKER like A

/ LINE~CKERTHE

FIG. 11,7, lxumple (49,,) lett and (49hl right

356 ROSS AND COOPER 11. LIKE SYNTAX 357

(50a) How should we get the onions to my sister? (SOb) Why don't we like MAIL your sister the onions? (SOc) *Why don't we MAl L your sister, like, the onions?

Similarly, (51c) is not a possible like-hopped version of (51b):

(5Ia) How can I get my aunt off of the dock? (Sib) Why don't you like PUSH your aunt off! (Sic) ·Why don't you PUSH your aunt, like, off!

Nonetheless, the prediction seems to be too strong, in some cases, Sentence (50a) can be replied to by (52a), and it seems to us as if(52b), which contains a pause-bounded like, can also be used, despite the fact that the focused element, MAIL. is not immediately followed by the pause-bounded ltke.

(52a) Why don't we like MAIL the onions to your sister? (52b) Why don't we MAIL the onions, like, to your sister?

,We have not found any satisfactory account of the difference between .(44d), ·(50c), and ·(5Ic), on the one hand, and the unexpectedly acceptable (52b), on the other.

There is one final factor that influences the operation of "Like"cHopping. Consider the (b) versions of (53) through (56), which are answers to the (a) versions. From all that we have said todate, we should expect like-hopped variants to exist. Nonetheless, the predicted (c) versions are alI blemished, toa greater or lesser extent.

(53a) Whose wallet is this? (53b) That's like BOB's wallet. (53c) *That's BOB's, like, wallet.? (54a) How heavy was that wallet? (54b) It was like UNEXPECTEDLY heavy. (54c) "It was UNEXPECTEDLY, like, heavy. (55a) How should we position the microphones with respect to the

bridge? (55b) Why don't you put them like UNDER the bridge. (55c) ?*Why don't you put them UNDER, like, the bridge.

'We have astcrisked (53c)only as a like-ho'pped answer to (53a), There arc circumstance, :""k' which the same Siring of words as (53c) can he used. with a pause after like, hut not bclorc. (,_ suggest that the speaker is not sure whether wollet is the correct word to use todescribe the "dJ steel and sandstone device that he knows Bob carries around his money in, We might punctuate this sentence as in (il. "That's Bob's like -wallct."We believethis type of "hesitation-lik c." to gl\ ~

it a name, to have different properties Irom hopped like. but we have not studied this question I" any detail.

(56a) How should we get the onions to my sister? [=.50aJ (56b) Why don't you like MAIL the onions to your sister. [=50b] (56c) ??Why don't you MAIL, like, the onions to your sister.

Note that the focused elements in these sentences are all on left branches. It is, thus, almost generalIy true that like may not be hopped over a

constituent, if the constituent that is hopped over opens a larger constituent. We have qualified with "almost" because, for reasons we do not understand, like can hop over a subject NP, despite the fact that the subject is on a left branch.

Another case where like seems not to be able to hop over a left branch is in coordinate structures:

.1 (57a) Who is similar to you? (57b) Like TOM and 1 are similar. (57c) P.TOM, like, and I are similar.

To account for (54) through (57), we introduce the following restriction.

(58) The Left- Branch Hopping Ban: "Like': Hopping is blocked if term 3 of (36) is a left branch of any node type other than S.

Armed with the observations concerning hopped likes. let us return to the LLBC. If we are correct in our claim that like can only hop over constituents, then it is most probable that when like precedes a constituent, it forms a structure such as that of (59) (see Fig. 11.8), to which "Like "Hopping will always be able to apply, unless AI is a left branch. After it has applied. the resulting structure will be as shown in (60) (see Fig. 11.9).ln this formulation. we assume arbitrarily that such "nodes" as COMMA (which signal the rules that assign intonation contours to produce high F, on both sides of like) are sisters of like.

Now we see that another way of formulating the relationship between like and the elements it is associated with is to use the relationship of C-command (Constituent command), introduced and argued for by Reinhart (1976). NOde

/\A,

A2FIG. 11,8. Example (59) like

A,

FIG.11.9. Example(60) ,~'"

=Q*m1=1~-

358 h~ _-, AND COOPER

s s ,

~ ~ NP, might appear N~ might appear

A ~ like NP2 NP2 CO...'" A like CO"'MA.

6 6 A DE"'ON A DE"'ON

FIG. 11.10. Example (63a) left and (63b) right

A C-commands node B in tree T if the next node up from A dominates B.· Given this notion, we can replace (26) and (27) by (61).

(61) The "Like" as a Cscommander Condition (LCC) Like must be adjacent to a constituent that C-commands the focused element.

To give an example of the LCC, it would allow either (62b) or (62c) to be used as answers to (2Ia), which is repeated for convenience as (62a).

(62a) What sort of thing might appear, if we say this spell? (62b) Like A DEMON might appear. (62c) A DEMON, like, might appear.

If (62b) has the structure shown in (63a), "Like't- Hopping will convert it into (63b) (see Fig. 11.10).

Note, however, that the LCC would also allow (64), see Fig. 11.11, to be the structure of (62b).

Like C-commands NP 1 in (64), because 8 1 dominates NP,. And (64) would' appear to be necessary, in order for "Like'<Hopping to be able to apply to produce (65), which is also a grammatical reply to (62a).

(65) A DEMON might appear, like.

It may even he possible to argue on intonational grounds for associating both (63a) and (64) with (62b): Note the possibility of pausing after lik» in (66).

'This relationship is. thus. the inverse of Klima's (1964) relationship in construction 11111,

11. LIKE SYNTAX 359

S,

lik~S2

I '6~mIght appear

FIG. 11.11. Example (64) A DEMON

N({\'~."V

F1G.11.12. Example (67) 6 I 'IA DEMON might appear' like

(66) Like, A DEMON might appear.

It may be that arguments can be found that this sentence has only structure (64), and the pauseless structure of (62b) has only (63a). We will leave this question open. For the present, note that if sentence (23b) has the structure of (67), see Fig. 11.12, the LCC would not correctly characterize it as a possible reply to (23a).

We noted earlier in (12a) that like cannot immediately precede the head noun of an NP. If we are to retain the LLC and yet prohibit like from occurring just before the head noun, we must add an ad hoc restriction prohibiting like in this environment.' We note that the restriction is particular to NPs, because like can occur in other situations between a modifier and its head, as in (68).

'Anothcr ad hoc restriction on like'« occurrence may hc needed to prohibit like before scntcncycomplements, as in

(i) ·We won't report like that you're sick, (ii) 'i'lI'1I arrange like for James to sleep upstairs.

""10 get f(iii) We tried like .. '. hack carlv, hut we were delayed.~ gettlllg .

whetherf

~(iv) I wonder like when I'll gel another ella nee.. where

360 ROSS AND COOPER

(68a) His proposal was almost like laughable. (68b) I nearly like strangled him.

Although we remain puzzled by the within-NP constraint on like, it is possible that this restriction may be understood in terms of speaking rate. We noted earlier that the occurrence of like is partly dependent on moderate to fast speech rates, and it might reasonably be argued that like is prohibited within an NP because NPs represent constituents that speakers typically utter as integral (cases of hes-itation represent a clear departure from this generality) rhythmic units. The insertion of like within the NP would break up the normal moderate-to-fast rate at which the NP could be uttered, accounting for the otherwise ad hoc restriction.

The prohibiting of like within simple NPs has a consequence for the analysis of prepositional phrases. Note that like can occur between a preposition and its head, as in (69), as well as before the entire prepositional phrase.

(69) I studied architecture ~n >"Chicago.

If prepositional phrases are analyzed as PP - P + NP (see Jackendoff, 1973), than like's occurrence between preposition and head requires no modification of the constraint prohibiting like within simple NPs. However. if prepositional phrases are viewed as NPs analyzed as NP-P + NP, then the restriction on like would need to be modified to prohibit like only within NPs that do not dominate NP.

LIKE AND ITS NEIGHBORS

Even

When one examines a wider range of lexical items in English. to try to find parallels to the behavior of like. even immediately presents itself. This obstreperous word has been the subject of scrutiny in a number of syntactic and semantic studies. among them Kuroda (1965), Horn (1969). Anderson (1972), Fauconnier (1975). Fraser (1971), and Jackendoff (1972).

Pauseless ..Even '' Possibly the most salient parallel to like is the fact that even can only be associated with a focused element. Thus. because the pronoun it cannot ever be stressed.s it cannot be preceded by even:

hExcept. for some reason. when it appears as a first conjunct:

. . { it and the vase} .(I) The table lipped over. but ? I d i both survived the fall. .tle vase an It

11. LIKE SYNTAX 361

(70) { *'t }We could wash even .t~at .

Thus, when we return to the three questions of (15a) through (17a), which are repeated for convenience as (7Ia), (71b), and (7Ic), respectively,

(7Ia) (71b) (71c)

Who is sending eggs to Marie? What are they sending to Marie? Who are they sending eggs to?

we find that by replacing like by even in (21), which contains three possible answers to (15), (16), and (17), grammaticality is preserved: cf. (72).

I . .1

1 ~I

~

(72a) (72b) (72c)

Even THEY are sending eggs to Marie. They are sending even EGGS to Marie. They are sending eggs to even MARIE.

.. ~

1 I

! i

In (22), we showed the results of repositioning like in (2Ia): in (73), we see an almost identical pattern ofgrammaticalities for the reorderings of the even in (718).

(73a) THEY are even sending eggs to Marie. (73b) *THEY are sending even eggs to Marie. (73c) *THEY are sending eggs even to Marie. (73d) *THEY are sending eggs to even Marie.

The one puzzling disparity is in the contrast between the fully grammatical (73a) and the controversial (22a): Although not all speakers always allow post-auxiliary like to be associated with a focused subject, this never seems to be objectionable in the case of even.

Proceeding with our comparison of like and even, we find that just as the unstarred sentences of (24) can serve as answers to question (16a) [= (71b»),so can the unstarred sentences of (74), which exhibit the results of systematically repositioning the even in (72b).

(74a) (74b) (74c) (74d) (74e) (741) (74g)

*Even they are sending EGGS to Marie. They even are sending EGGS to Marie. They are even sending EGGS to Marie. They are sending even EGGS to Marie. [= (72b» *They are sending EGGS even to Marie.' *They are sending EGGS to even Marie. *They are sending EGGS to Marie even.'

'These examplesare to be read with pauseless evens. The behaviorof pause-boundedeven will be discussed later.

362 ROSS AND COOPER

Finally. just as (74) parallels (24). (75) parallels (25):,if one repositions even in (72c). the following results:

(75a) ·Even they are sending eggs to MARIE. (75b) They even are sending eggs to MARIE. (75c) They are even sending eggs to MARIE. (75d) ·They are sending even eggs to MARIE. (75e) They are sending eggs even to MARIE. (75t) They are sending eggs to even MARIE. [= (72c)] (75g) They are sending eggs to MARIE even)

The only important differences between (24) and (74), and (25) and (75), are in the (a) and (b) versions: Although /ike is good before subjects and bad after them, even shows the opposite behavior. This complementary behavior can be described well by either of the following transformational rules:

(76) "Even" In X - [even [ NP - V Y ] ] - Z

s s s s 2 3 4 5 OBL

I 0 3 + 2 4 5 ~

(77) "Like" Out X - [ NP - like - V Y ] - Z

s s 2 '3 4 5 OBL

3# [2 0 4 ] 5 ~

s 5

The question that confronts us here is: Which one of these candidates for transformation hood is the correct rule for English? Should we start out by generating both like and even sentence-initially, and then move eve,n. rightwards, as in (76), or should we instead say that like and even start out to the right of subjects, with like being moved obligatorily to the left, as in (77)7

There is some evidence that favors the adoption of an "Even "In analysis. but as it is complex, involving the structure of the auxiliary. we will defer its presentation until the section The Constituent Structure of the English Auxiliary.

Having argued for the conclusion that like and even are identical in distribution. except for the difference produced by one of the rules in (76) and (77), let us now attack this conclusion. for in fact, there are many cases where like is possible, but even is not. Some are cited in (78) through (86).

(78) Before postnominal modifiers. full or reduced:

11. LIKE SYNTAX 3'63

(78a) The kids {.like } who were in THE LIVING ROOM were quiet. even

(78b) The kids {?like } in THE LIVING ROOM were quiet. .even

' like } .(78c) Somebody who was INSANE must have attacked this { even lemon meringue pie.

like } .(78d) Somebody. who was INSANE must have attacked this{ even lemon meringue pie.

like } .(78e) Somebody. INSANE must have attacked this lemon { even meringue pie.! .

(79) One "even" per surface clause. multiple likes:? (79a) [Only hard-core male supremacists would stoop to reading girlie

mags. But there appear to be some sexist pigs even among the most intelligent members of our group. Bob, whose IQ is 120, reads even Hustler. And Tony's IQ is higher: 159.] ·And even Tony reads even Hustler.

(79b) Q: What kind of bike was Hoboridin', man? A: Like he was like riding like a huge BSA, like.

(80) Before contrastively stressed auxiliary complexes:

(80a) It {at. like IO} IS possible that we'll win. 70even .

like } .(80b) James Bond. was NOT working for Smersh. { even (81) Before the focus of a cleft or pseudo-cleft sentence:

like }(8Ia) It was. PETER that I meant. { even like 'l(81b) (The one) who I meant was. I PETER.{ even.

(82) Before universally quantified NPs:

'Note that whereas the ungrarnmaticality of(78a) and (78d) could be attributed to a failure to undergo the obligatory rule of"Even"/11. no such account can explain the "?"of (78b) and the' of (78e).

"This important observation. which is due to Kuroda (1965). is also discussed insightfully by Anderson (1972).

"One of the authors does not have this contrast in his (sloppy) speech.

11. LIKE SYNTAX 365364 ROSS AND COOPER

J'k (82a) We are fingerprinting.(* I e

~ even

EACH ) 'EVERY STUDENT {

}'(ANY STUDENT(S) " NO STUDENTS ,

,ALL STUDENTS,

(82b) { ;~~:n} when were you in East Berlin?

(83) After degree modifiers:

like }(83a) Jefferson was AFRAID of the mouse, { even

.very')

'extremely { { like }(83b) Jefferson was (somewhat, * AFRAI D of the mouse.

. (evenquite ,

,etc.

(84) Before (most? all?) indefinite quantifiers: (84a) The Feds nabbed SOME 'protesters, and the local cops

busted { like} {SEVERAL}*even MANY

(84b) We don't subscribe to MANY magazines, but

. { like}we do subscribe to * SOME. even

(85) Before sentence adverbs:

. {like}(85a) It IS POSSIBLE that he'll be back at 5. even

Like } ,(85b)' { *Even POSSIBLY, hell be back at 5.

(86) Before adverbs that do not end in -~l':

(86a) Sarah picked me {*Iike }RIGHT UP. even

(86b) They have been working

(86c) He might show up here

{ like } STRENUOUSLY even

j like} HARDt ?*even

j like} FREQUENTLYt even

like } OFTEN{ ?*even

With this impressive array of differences (most, as far as we know, unexplained by previous analyses of even), one might be tempted to generalize as in (87).

(87) Wherever even is possible, like is, but not the converse.

But (87) is also incorrect, as shown by (88).

?*Iike } He realized' that HE was unpopular.(88a) { even

(88b) The mail {*?Iike } Jimmy CARTER gets is opened..even

*Iike } (88c) TOM must have gone by now, and JANE may have,{ even too.

At present, we must be content to state that, while (87) cannot be maintained, like is generally freer than even.

Pause-Bounded "Even". In the immediately preceding section, we showed how pauseless even exhibits strong parallels to pauseless like, with respect to the structural configurations in which an even can appear when it is not adjacent to the focused element. Interestingly, the behavior of pause­bounded even also shows an exact parallel to the behavior of pause-bounded like. This parallel can be accounted for by the account of even provided hy Anderson (1972), (see later discussion), which has the same empirical consequences as our LCC restriction for like.

The sentences in (35) were the like-hopped variants of the sentences in (21). The sentences in (72), which are the result of replacing the likes in (21) by evens, can also undergo hopping, with the sentences in (89) being the result.

(89a) THEY, even, are sending eggs to Marie. (89b) They are sending EGGS, even, to Marie. (89c) They are sending eggs to MARIE, even.

Just as we argued that pause-bounded like must immediately follow a constituent containing the focused element, so we argue for pause-hounded even. Above, in *(44d), this condition was violated for-pause-bounded like; in *(90d), a corresponding ungrammaticality has arisen.

(90a) Was EVER YBODY working on Greek? (90b) Yeah-even BOB was working on Greek.

'~~~~~P;i' .~t=~:!:~I:::'..

366 ROSS AND COOPER 11, LIKE SYNTAX 367

(90c) Yeah-BOB. even, was working on Greek. (96c) They are sending eggs to {???nlY} MARIE.

" (90d) ·Yeah-BOB was, even, working on Greek. . rjust I,

" ~ And just as pause-bounded like can appear sentence-finally, in (45), so we .And just as we can find even to the left of MARIE in (75), so we also find 011(1' ,~ find a final pause-bounded even in (91). and just in the same kinds of environments: cf. (86).

(91) Y~ah-BOB was working on Greek, even. (97a) • {~u~li } they are sending eggs to MARl E.

And just as hopping like around left branches of constituents other than S produces varyingly unhappy results (such as ·[53c), ·[54c), ?·[55c) and (97b) TheY{jOU~~} are sending eggs to MARIE. ??[56c), so does hopping even in similar sentences: cf. (92c) through (95c)

. \, below. (97c) They are {ju~~}sending eggs to MARIE.

(92a) Did they weigh EVERYBODY'S wallet? ~Ii I(92b) Yeah-they weighed even BOB'S wallet, i (97d) ·They are Sending{ju~~} eggs to MARIE.(92c) ·Yeah-they weighed BOB'S, even, wallet.

(93a) Was the wallet VERY heavy? I (93b) Yeah-it was even ALARMINGLY heavy. (97e) They are sending eggs{ju~~} to MARIE.1(93c) ·Yeah:....-it was ALARMINGLY, even, heavy. (94a) Should we put microphones EVERYWHERE-ALL AROUND

However, when we try to hop these words, we find radically different the bridge?

patterns from those we encountered with /ike and even. First of all, it appears I(94b) Yeah-why don't you put them even UNDER the bridge. that there are no contexts whatsoever in whichjust can hop: compare ·(97), as (94c) ?·Yeah-why don't you put them UNDER, even, the bridge. 1

r i answers to (71), with (96). "I (95a) Should we use ANY MEANS to get the onions to my sister? ~' :

(95b) Yeah-you might even MAIL the onions to your sister. (98ai) ·THEY (,) just (,) are sending eggs to Marie. (95c) ??Yeah-you might MAIL, even, the onions to your sister.

(98aii) ·THEY are just sending eggs to Marie. (98b) ·They are sending EGGS (,) just (,) to Marie. It, thus, seems fairly clear to us that our rule (36) is stated too narrowly: It (98c) ·They are sending eggs to MARIE (,) just. must be generalized to apply to even as well as to like. However, we are not

much nearer to understanding like-Hopping merely because we can show that A further indication that just never hops is the fact that it never occurs

~l one other similar item hops. Indeed, if being associated with, the focused

sentence-finally.element is the only criterion of similarity, then we might expect any element that shares this property with like and even to hop similarly. Not so. Let us

~.fi (99) ·THEY are sending eggs to Marie, just. widen our field of view andexamine the behavior of some other focus-linked i',~ items: only and just, on the one hand, and also and too on the other.

In general (and for reasons we do not understand)just is much more limited f:: in its distribution than only, as was already apparent in (96c). A further point "011(1''' and "Just." The words 011(1' andjl/st parallel the behavior of like

of difference.is the fact that there are contexts in which a pause-bounded 01111'~:, and even in many ways, as can be seen by comparing (96), used as an answer to can appear sentence-initially, with roughly the sense of "There's only one t (71), with (72).

~ .. (\;, thing: X." ~)

(96a) {~u~~}THEY are sending eggs to Marie. ~~j I } {he wasn't in his room last night. }(?'

.... , (100) l.J 0

n i ' be careful where you put the H2S0 4• ~~ t (96b) They are sendin g{ ?nIY} EGGS to Marie . us where can we get some Kool-aid to mix with it? Justt·:

l~~;t%~~.:-:.

368 3 AND COOPER

As is evident from the asterisk in (100). this isanother case where only but not just can appear. The following generalization seems to obtain:

(101) Wherever just can appear with the sense of only that is described in Horn (1969), only can also appear.

An exception to (10I) appears in (102),whichshows that onlyjusl can appear before imperatives, although not with the most common meaning of onlyas described in Horn (1969).

*0 I } 1get lost }(102) { JU~! b~ quick about it !

give me that

Let us turn now to an examination of the hopping behavior of only. First of all, the most typical intonation for a post-focus only has no pauses, as can be seen in (103), which corresponds to (96).

I 2 I

(103a) THEY ONLY 1 1 ( are sending eggs to Marie.

??THEY, ONLY,

I,2 I I. EGGS ONLY .(103b) They are sending I I to Mane.

??EGGS, ONLY,

. MARIE ONLYI 2 I(103c) They are sending eggs to I II .

??MARIE, ONLY

That is, the standard hopped 011(1' is pausetess. This may beconnected with the fact that hopping only around a left branch always produces solid stars. whereas like and even hop around some left branches without IOl.Illy destroying the sentence. Thus, compare the (c) versions of'(104) through (1ll7) below with the (c) versions of (92) through (95).

(104a) Did they weigh EVERYBODY'S wallet? (104b) No, they weighed only HOS'S wallet. (f04c) *No. they weighed BOB'S only wallet. (105a) Was the wallet VERY heavy'! (105b) No, it was only SOMEWHAT heavy. (105c) *No. it was S;OMEWHAT only heavy.

11. LIKE SYNTAX 369

(106a) Should we put microphones EVERYWHERE-ALL AROUND the bridge?

(106b) No-why don't you put them only UNDER the bridge. (106c) *No--why don't you put them UNDER only the bridge. (107a) Should we use ANY MEANS to get the onions to my sister'! (107b) No-why don't you only MAIL the onions to your sister. (107c) *No-why don't you MAIL only the onions to your sister.

Also and Too

Having compared the hopping behavior of like and even with that of only and just, let us turn to the focus-bound elements also and too. When these words are associated with subjects, as in answers to a question such as (108), their syntactic behavior is as shown in (109).11

(108) Were only THE DOSSASES sending eggs to Marie? 2 I

(109ai) No, ALSO THE PILL/NGS were sending eggs to Marie. I 2

(109aii) *No, THE PILLINGS ALSO were sending eggs to Marie. 2 . I

(109aiii) No, THE PILLINGS ALSO were sending eggs to Marie. 2 I

(I09aiv) No, THE PILL/NOS were ALSO sending eggs to Marie. 2 J

(l09av) *No, THE PILLINGS were sending ALSO eggs to Marie. 2 I

(109avi) No, THE PILL/NOS were sending eggs to Marie ALSO. (109bi) "No, TOO THE PILLINGS were sending eggs to Marie.

I 2 (109bii) *No, THE PILLINGS TOO were sending eggs to Marie.

I I (I09biii) 'INa, THE I'ILLlNGS TOO were sending eggs to Marie. (I09biv) *No, THE PILLINGS were TOO sending eggs to Marie. (109bv) **No, THE PH_LINGS were sending TOO eggs to Marie.

2 I (109bvi) No, THE PILLINGS were sending eggs to Marie TOO.

The salient points about also and too include the following.

I/\Ve mark (he location of primary and secondary sentence stress ill SOTTle (11' thl,,' following examples with a superscripted "'I" and "2" above the appropriate syllables, Thi.'\notntiou is. thus, ill the spirit of the notation of Chomsky and Halle (196&).

370 ROSS AND COOPER

(I lOa) In contrast with like. even. just and also. 100 may never precede the focused element. [cf. *(109bi)J

(I lOb) In contrast with like and even, post-focus only and post-focus also and 100 must carry main sentence stress. [cf. the contrasts in *( lO9aii) and (I09aiii) and *( I09bii) and (109biii).]

(IlOc) Also can appear adjacent to the focused element, in post-tensed­auxiliary position or sentence-finally. Too can appear only immediately after the focused element, or sentence-finally.

When we extend our investigation to answers to questions such as (III) and (112), in which the focused constituents are direct and indirect objects, respectively, it appears that the generalizations stated in (110) stand. Sentence (I13) answers (III), and (1I4) answers (112).

(Ill) Were they sending only EGGS to Marie? (112) Were they sending eggs to only MARIE?

2 I (II3ai) ?No, ALSO they were sending BANANAS to Marie.

2 I (I I3aii) No, they ALSO were sending BANANAS to Marie.

2 I (I13aiii) No, they were ALSO sending BANANAS to Marie.

2 I (113aiv) ?No, they were sending ALSO BANANAS to Marie.

I 2 (113av) *No, they were sending BANANAS ALSO to Marie.

2 I (l13avi) ?No, they were sending BANANAS ALSO to Marie.

2 I (113avii) No, they were sending BANANAS to Marie ALSO.

(ll3bi) **No. TOO they were sending BANANAS to Marie. (l13bii) **No. they TOO were sending BANANAS to Marie. (113biii) **No. they were sending TOO BANANAS to Marie.

2 I (113biv) No. they were sending BANANAS TOO to Marie.

2 I (113bv) No, they were sending BANANAS to Marie TOO.

2 I (114ai) ?No. ALSO they were sending eggs to MARIE:

2 I (l14aii) No, they ALSO were sending eggs to MARIE.

2 I (114aiii) No. they were ALSO sending eggs to MARIE.

11. LIKE SYNTAX 371

2 I (114aiv) *No, they were sending ALSO eggs to MARIE.

2 I (114av) No, they were sending eggs ALSO to MARIE.

2 I (114avi) ??No, they were sending eggs to ALSO MARIE.

I 2 (114avii) *No, they were sending eggs to MARIE ALSO.

2 I (114aviii) No, they were sending eggs to MARIE ALSO.

(114bi) "No, TOO they were sending eggs to MARIE. (1I4bii) **No, they TOO were sending eggs to MARIE.

(114biii) **No, they were TOO sending eggs to MARIE. (114biv) **No, they were sending TOO eggs to MARIE. (114bv) "No, they were sending eggs TOO to Marie.

(114bvi) "No, they were sending eggs to TOO Marie. I 2

(1l4bvii) *No, they were sending eggs to MARIE TOO. 2 I

(1I4bviii) No, they were sending eggs to MARIE TOO.

A few additional comments about (109), (113), and (114) are in order here. First, the ungrammaticality of *(109av), **(109bv), and *(l14aiv) shows that also (and 100) are similar to like and even in obeying the LCe. Furthermore. the ungrammaticality of *(97d) shows thaton~Y and jus I are also subject to this constraint. As far as we know, therefore, the generalization stated in (115) can be maintained.

(115) All focus-linked elements obey the LCe.

Second, we note from examples such as ?(113ai) and ?(114ai) that. in contrast with only and just, sentence-initial also is not totally excluded. Because only andjusl must, like even. be ruled out in sentence-initial position [cf. *(97a)], we will have to generalize our rule of"Even "In so that only and just will also be moved obligatorily to post-subject position; and a condition will need to be added that specifies that such a repositioning rightwards is preferable for also, although not mandatory.'?

Third, for reasons that remain opaque to us, of all the focus-linked elements that can ever precede the focus, also most strongly resists being inserted between a preposition and an NP [cf. ??(114avi) and (116)].

12J'''lcrnatively. we can generalize "Like" Out Sf) that it applies obligatorily us lik» and half­heartedly to also. Both solutions look messy. We find it hard to invest much in either.

"r. ,:'~ ~,~,.

11. LIKE SYNTAX 37~372 I AND COOPER

like even

(116) I was talking to { only \ MORGAN about that. ?just

?*also

Let us return now to the issue of the stress on also. Examining the (a) versions of (109), (113), and (114), a clear pattern emerges:

(117) Regardless of the relative order of also and the focused element, the rightmost of these two bears the primary sentence stress, with the leftmost bearing the second highest stress in the sentence.

This is a particularly interesting generalization, for it is unstatable as a phonological rule in any framework that we know of. Therefore, rather than trying to construct a totally new type of phonological rule, we tentatively propose a syntactic solution.

We assume that the stress dependence of also and the focused element isan indication that they once were members of the same phrase. If they are reordered by a generalized rule of Intensifier-Hopping, which is ordered before the Nuclear Stress Rule;'? then the rightmost element will receive the highest stress.

So far, we have succeeded only in accounting for the stress differences of (l09ai)/(I09aiii), ?(I 13aiv)/(I13avi), and ??(l14avi)/(l14aviii). But what about the fact that in (113ai), also has [2 Stress], while it bears [I Stress] in (l l Javii)? These sentences are synonymous, as indeed all sentences in (113) are. What we propose to do is to account for this synonymy by deriving the sentences in (113) in which also precedes bananas from (l lJaiv), and (113avii), in which it follows bananas, from (l l Javi), by the following rule:

(118) "Also" Climb W - [BX - [ also - A ] - YB] - Z

I 2 3 4 5 6 OPT ~

13# [B2 o 4 5~ 6

This rule will Chomsky-adjoin an also which precedes its focus to the lett of any higher node.!" The rule must follow the two rules of Intensifier­Hopping and the Nuclear Stress Rule, as shown in (119).

"I'or a description oj IllIS pt nccsx. sec Chomsky alld Halle (I9!>R). "he suggestion that this rule should precede certain syntactic operations appeared firS! ill Bresnan (1971).

"Ohviollsly. Ihis is only half of the required rule. which must also raise alsos that follow their focus and adjoin them to the right oj any higher node, That is. rule (102) should be rewriucn as a mirror-image rule (cf. l.angackcr [1969] for discussion). .

(\ 19) Rule ordering A. Intensifier-Hopping (OPT) B. Nuclear Stress Rule C. "Also" Climb (OPT)

Given this system, (l13aiii) and (113avii) will be derived as in (\20) and (121), respectively. IS

I I (120a) No, they were [sending ALSO BANANAS to Marie] NSR

~2 I (l20b) No, they were [sending ALSO BANANAS to Marie] "Also"

Climb 2 I ~

(120c) No, they were [ALSO [sending BANANAS to Marie]] I I lntensifier­

(12Ia) No, [they were sending ALSO BANANAS to Marie] HOPfJin~

S S ~

I I NSR (121b) No, [they were sending BANANAS ALSO to Marie] ~

2 I "Also" (121c) No, [they were sending BANANAS ALSO to Marie] Climb

I ~

( 121d) No, [[they were sending BANANAS to Marie] ALSO]

The other derivations in (109), (1/3) and (114) will proceed similarly. We return to the rules in (119) below in subsection Conclusions on "Like"

and Intensifiers, where we propose extending them to handle much of the distribution of intensifiers. First, however, let us examine the behavior of also when it is hopped around left branches.

Sentences (122) through (126) show that hopping also around left branches results in sentences that are totally unacceptable, as was the case with only,

(122a) Did they weigh only ZEK E'S wallet'! (122b) No, they weighed also BOWS wallet. (122c) *No, they weighed BOB'S also wallet. (123a) Do you mug only EXPENSIVELY dressed men? (123b) No, I mug also POORLY dressed men. (l23c) *No. I mug POORLY also dressed men. (124a) Should we put microphones only ON the bridge?

"We have drawn in the brackets surrounding the constituent to which olsoattaches. although wehave not labeled them. for the issue as to whether they should he called Ss or VI's is irrelevant here.

l~' 374 ROSS AND COOPER 11. LIKE SYNTAX 37Sr:~ ....~.

~~. ,~~

(124b) No-why don't you put them also UNDER the bridge. Condition: If 3 + 4 is the NP object of a preposition. the rule j, (I 24c) *No·-why don't you put them UNDER also the bridge. obligatory if3 =also, and is preferred if3 =jus/.[ Ct. (125a) Should we only CARRY the onions to my sister? (125b) No, why don't you also MAIL the onions to your sister.

MAIL ALSO.I .I(I 25c) !*N 0, why don't you I 2 the onions to your sister. I 2

MAIL ALSO I I

'I; Summary ofrules for intensifiers. Let us summarize the discussion of the :Ii section "Like" and its Neighbors up to this point. We have found the '\ folIowingsimilarities in the behavior of intensifiers, i.e., offocus-Iinked items:

(I26a) AlI obey the LCe. (126b) AIl (except just) can appear to the right of the focused element

(too can only appear there). (126c) All obey the LBHB.

We noted in (117) a generalization about the interaction of a/so and the location of primary sentence stress, suggesting that a syntactic treatment, involving the ordering of the Nuclear Stress Rule before a syntactic rule of "Also" Climb, was preferable to the creation of a new type of phonological rule. This leavesus in a rather redundant situation with respect to the LCCfor the following reason. The rule of "A lso'' Climb makes it unnecessary to state the LCC for also, for the output of any sentence with a climbed also will necessarily obey the LCe. Thus, for a/so, the LCC can be dispensed with. in favor of the rule of "Also" Climb.

The trading relationship betweenthe LCC and "A lso" Climb in the caseof also immediately suggests that the same type of relationship may obtain in the case of the other intensifiers, given their widespread similarities in distribution. That is, for the other intensifiers, weare naturally led to look to see·ifthere is any independent reason to choose the LCC over a solution that would involve climbing intensifiers other than also.

As far as we have been able to ascertain, no such evidence exists. And because the stress facts noted in (117) above can be accounted for, given a climbing solution, but cannot be accounted for merely by the Lee. we conclude that the LCC can be dispensed with entirely, in favor of the following rule. which should also be construed as a mirror-image rule. as discussed in Footnote 14.

(127) Intensifier Climb W - [BX - [AIntensifier - AA] Y -] Z

I 2 3 4 5 6 OPT I 3# [B 2 0 4 5 6 ~

B]

( 1(6)]

The variables in this rule deserve much more study than we have been able to carry out. Such questions as the following arise:

(128a) Is Intensifier Climb a bounded rule?That is,can intensifiersclimb out of clauses?

(l28b) Is Intensifier Climb subject to constraints on movement rules, such as those proposed by Ross (1967)?

(128c) Are there asymmetries in the climbing behavior of intensifiersthat are on left branches vis-a-vis those that are on right branches? ,I

"

j Our answer to the first question isa blunt "yes and no." Intensifiers seemto

be able to climb out of alI (?) nonfinite complements and out of some tensed clauses: the (b) and (c) versions of (129) through (131) are synonymous

.~ answers to the (a) versions: ,j

l

~l want I (129a) Did he I hbegin to translate anything else into French? I ope

. . etc.

Iwanted be an .

( l29b) Yeah, he hoged I to translate even YOUR LETTERS into French. t p

I e c.

wantedI ( l29c) Yeah, he even 1~~~~~ to translate YOUR LETTERS into

French. t e c.

avoid

(130a) Did they stloP l kicking only MORTON? p an on Ietc.

avoided I stopped kicking also YOUR UNCLE. (l30b) No, they planned on etc.

avoided I (l30c) No, they also I stlopped kicking YOUR UNCLE.

P anne d on etc.

,/

376 f, AND COOPER

think jbelieve(131 a) Do the bosses {h that only ZACK was at the rally? ope say

thinkbelieveI I?also OSCAR that(13Ib) No, they was at the rally. hope OSCAR also say

think believe

(131c) No, they also that OSCAR was at the rally. hope say

However, it does not seem possible to climb intensifiers out of [active clauses, with strong [actives likesurprised blocking such a putative reordering more clearly than do weak factives like know. Thus, whereas (l32a) and (132b) are probably impossible to read as synonymous, (l33a) and (133b) seem to us to definitely not be synonymous.

(132a) Do the police know that Dr. Hunger also IRRADIATES us'! (I 32b) Do the police also know that Dr. Hunger IRRADIATES US'!

? [¢ (132a)]

(l33a) Are the police surprised that Dr. Hunger also IRRADIATES us? (133b) Are the police also surprised that Dr. Hunger IRRADIATES

us? [¢ (l12a)]

Thus, it appears to us that Intensifier Climb will have to be made subject to the type of nondiscrete constraints on variables that are discussed in Ross (1975), where a number of other processes are cited that behave differentially with' respect to complements of verbs such as think as opposed to complements of verbs such as surprise.

The answer to the second question in (123) appears to us to be extremely unclear at present, despite the fact that Anderson (1972)cited such Complex­NP-constraint violating cases as (134) [= Anderson's (9a)], (135) [= Ander­son's (17a)), and (135) [= Anderson's (10)]

(134) You can do a lot of things with Skrunkies: 1even know a guy who SMOKES them.

(135) I even included a problem that FRESHMEN could solve. (136a) John even has the idea that HE is tall fora Watusi. (136b) John even has the idea that he is tall for a WATUSI.

11. LIKE SYNTAX 37',

and such cases as (137) [= Anderson's (12)], for which even would have to climb out of a sentential subject.

(137a) J ones' wife considers him absolutely honest; I imagine his even THINKING such a thing would amaze her.

(137b) Jones' wife considers him completely honest; I even imagine his THINKING such a thing would amaze her.

Although we agree in general with Anderson's (1972) grammaticality judgments here (although we find (l37b) more' dubious than he does). we note that have the idea that does not make very strong islands: whereas questioning out of a following S is difficult (cf, (138a», relativizing often seems acceptable [cf. (138)]:

(138a) ?What does John have the idea that he should build? (138b) The problems that John has the the idea that he can solve are all

exceedingly trivial anyway.

Thus, (136) does not provide solid evidence that the Complex NP Constraint should be viewedas having been violated. Nonetheless, even cannot climb out of a sentential subject, in (I 39b), (which corresponds to (139a), except that the even has made a similar climb), as the Sentential Subject Constraint would predict.

(139a) It's likely that there will be pictures of lots of planets; I'd say that there being a picture of even PLUTO is possible.

(139b) ?*It's likely that there will be pictures of lots of planets; I'd even say that there being a picture of PLUTO is possible.

Anderson's (1972) strongest cases, then, are (134)and (135). We agree that, if even leaves the relative clauses in these cases, vis Intensifier Climb, it will be violating the Complex NP Constraint. Our feeling, however, is that (134)and (135) may be atypical, for note that the (b) versions of sentences belowshould all be variants of the corresponding (a) versions, if the Complex NP Constraint is never involved in the relationship between even and its focus.

(140a) He ridiculed your hypothesis that we could even DETECT this radiation.

(140b) He even ridiculed your hypothesis that we could DETECT this radiation. [a ¢ b]

(14Ia) This compound dissolved the plastic that even THE AQUA REGIS left unblemished.

(14Ib) This compound even dissolved the plastic that THE AQUA ?

REGIS left unblemished [a ¢ b).

378 ROSS AND COOPER

Thus, we have not been able to ascertain whether even can or cannot generally climb out of complex NPs, It does, however, seem to obey the Coordinate Structure Constraint wholeheartedly:

(142a) I'll invite both Beth and Marge, and even TOMMY, to the party. (142b) *1'11 even invite both Beth and Marge, and TOMMY, to the party.

In sum, then, the rule of Intensifier Climb presents a mixed picture with respect to constraints on variables. Our present belief is that the theory

··1 advanced by Ross (1967), which sought, basically, to develop only constraints that all rules of all languages would abide by, is too inflexible. Rather, there appear to be stronger and weaker constraints and stronger and weaker rules. The Complex NP Constraint is a weaker constraint than the Coordinate Structure Constraint; and the rule of Question Formation is a weaker rule than the rule of Relative Clause Formation (which accounts for the contrast in (117) above). 16 The rule of Intensifier Climb appears to be even stronger than the rule of Relative Clause Formation, but riot so strong as to be absolutely constraintless, as Anderson (1972) seems to suggest.

Turning briefly to the third question in (128), we can see immediately that, although it is possible for intensifiers to move leftwards across nonfinite clause boundaries [cf. (129c) and (l30c), and maybe (137b)], they appear not to be' able to move across right-clause boundaries: Thus, (143a) cannot become (143b).

(143a) Your even SUGGESTING it to Doris was stupid. (143b) Your SUGGESTING it to Doris was stupid, even.

Thus, although it is not clear what sort of constraints should be imposed on the variable in Term 2 of rule (127), it does seem that this rule obeys the Right Roof Constraint that was proposed by Ross (1967): An intensifier that climbs from a right branch must command any nodes it is to climb over.

THE CONSTITUENT STRUCTURE OF THE ENGLISH AUXILIARY

"Even" In vs. "Like" Out

Let us now examine the consequences of our rule of Intensifier Climb for the analysis of auxiliary verbs. Consider (144a), a question that has a maximally complex auxiliary, and a possible answer, (144b):

"For some discussion of a nondiscretetreatment of constraints on variables.see Ross(, 975)

11. LIKE SYNTAX 379

(144a) Could Fay Wray have been being gripped by only RODAN? (144b) No, she could have been being gripped by even KONG.

If we now try to climb even in (144b), we find that it can attach to the prepositional phrase by KONG, and also that it can appear to the left of gripped, or to the left of any of the preceding auxiliaries,

(145a) She even could have been being gripped by KONG. (145b) She could even have been being gripped by KONG. (145c) She could have even been being gripped by KONG. (145d) She could have been even being gripped by KONG. (145e) She could have been being even gripped by KONG. (145f) She could have been being gripped even by KONG.

If the rule of Intensifier Climb is correct, these facts require that auxiliaries be assigned a right-branching structure such as that shown in (146), Fig. 11.13.

j'/\z~ "".,As... v

I"'''', "A I-­. " I

been

54A s

V/4 A beinQ vIs )~

gripped PI NIPI

by KONG

FIG. 11.13. Example (146)

..:rc-­;'IW.~~~·~

380 JSS AND COOPER

S,

..»>: 'I l' ~ She could hove been being gripped by KONG FIG.11.14. Example(147)

We need not be overly concerned with certain details yet. In particular, the Intensifier Climb analysis is compatible with a structure with the nest of right­

Ii branching nodes all being sentences, as in (146), or with them all being VPS.17 It

We have designated the right sister of the subject NP in (146) with the 1 cryptic emblem "?", because we wish to argue that, in fact, there should be no

node dominating the sub-string starting with could and ending with KONG. II Rather, we wish to maintain a ternary-branching analysis of SI, as in (147),

'I Fig. 11.14.

The reason that we find this preferable to the binary analysis of'( 146) has to do with the behavior of like, which seems to reveal more of the structure of auxiliaries than does the behavior of the other intensifiers. And we find, as

11 possible answers to (148a), a large number of the family of sentences that are :1, abbreviated by the parentheses in (148b): \; II (148a) Who could Fay Wray have been being gripped by?

',; (148b) (Like) she (*like) could (like) have (like) been (like) being (like)

gripped (like) by (like) KONG.

I The fact that like is only impossible before could in (148b) would follow from our formulation of Intensifier Climb, if (148b) had the structure shown in (149), Fig. 11.15.

We do not wish to claim that all of these likes must have originated from a copying process that would proliferate a single like that isadjacentto KONG

l. in remote structure. This could be the case here; but is arguably not the case in other multiple-like sentences. For the present discussion, let us restrict our

t

f attention to the variants of (l48b) that have only one like. which we will assume to have been moved, by Intensifier Climb. from being under its source node. the circled N P' of (149), to being Chomsky-adjoined to any of the six higher nodes, as shown in (149). I

Note that if there were some node dominating the whole string from could to KONG, namely, the node."?" of( 124), we wouldpredict that like could also attach to "?", which would yield the asterisked variant of (l48b),

pror previous treatments 01' auxiliaries in which they are. analyzed as right-hr;lIlchill~

constructions, see Ross (1~6~) andKeyser and 1'0,t,,1 (1976).

,/~\ (II~.

.> v: ~:\'2 NP '\ "

1 IdIi (\ik'~AI52.he l*likel cou

2 'V ..

have 5 3 I IIike'A

V3 :/\

I lI;k'el 54 been . A

/ ~\ I. (like) 55

being A

/ -, V5 :/\

I r PPgrippe. d (Iikel/\

@) p '\

byI (like I" ~I'" KOI'l(,

FIG. 11.15. Example (14~1

i~~_~jJP;~1~;:'·;~i '~J!.!?J~'1!~~L:~i~~~~;<....\~~, .

I

381

382 ROSS AND COOPER 11. LIKE SYNTAX 383

Hif:

Let us return to (144b) for a moment, to note that one place that even cannot appear is to left of the whole sentence:

(150) *Even she could have been being gripped by KONG.

Thus, although like can precede subjects, but not follow them, even (and the other intensifiers) have the opposite restriction.

We noted earlier in the subsection Pauseless "Even" that this situation could be accounted 'for in either of two ways: by postulating that the situation with even was basic, positioning like initially after the subject, and obligatorily moving it leftwards (by "Like" Out) analysis, or by postulating that the situation with like was basic, with an obligatory rule of "Even" In.

The following puzzle suggests to us that this latter course is correct; What, in a "like" Out analysis, would rule out *(150)? Evidently, it would have to be some condition on Intensifier Climb, something to the effect of (151).

(151) Condition: if 2 + 3 + 4 + 5 =S, and if 3 =even, only, just and (weakly) also; the rule blocks.

That is, even, only, just, etc. would have to be prevented from being attached to S.

There are, however, sentences that seem to indicate clearly that (15I) is too strong and that even must be able to be attached to S. Consider (152).

(152) It is possible even that she could have been being gripped by KONG.

Disregarding the question at issue at present, namely, what the structure of the complement of (152) is, the structure of the matrix clause is generally agreed to be representable as in (I 53a) or (I 53b), Fig. 11.16 (where the issue as to ·whether S2 is part of the VP of S\ is not germane to the present discussion). That is, it is agreed that what follows possible in (152) is a sentence, but not an NP.18

But, if (152) shows that even can indeed be Intensifier Climhed ! 0 he Chomsky-adjoined to S, then what explains the difference between (152) and the ungrammatical *(150)? Our hypothesis is that the cornplerncntizer is responsible. Note that if that is deleted in (152), as is generally permitted with

"The only theory of complementation with which we are familiar that would claim Ihat the tlrat-clause of (152) retains its NP status is the theory of relational grammar that is ptcscntl, under development by Perlmutter and Postal (1978). In this theory. it would be claimed t hat the that-clause is an NP. a chorneur. that has been produced by the insertion of the dummy ir in subject position. Until we are more familiar with the details of this theory. in particular with the evidence for the claim that extra posed clauses must retain their NP status, we continue in the familiar assumption that extra posed clauses are only dominated by S.

5,

NP

6 52

,

il

51

.>. "I ~ I .. i. Ii i\!

FIG. 11.16. Example (l53a) top and (I53b) bottom possible ~

adjectives such espossible, afraid. sure, lucky, likely, etc. [see( 154)], the result is ungrammatical [cf. *(155a)], unless even is moved after the subject, as in (155b).

It's possible I'm afraid

(154) {We're sure ~ (that) you'll survive a week in Las Vegas. It's lucky It isn't likely

(155a) *It is possible even she could have been being gripped by KONG. (l55b) It is possible she even could have been being gripped by KONG.

A related case shows the importance of cornplementizers for the formal statement of the rule that moves even rightward. Consider the following sentences, which indicate that the direct object NP of report can be modified by even, as we would expect:

(156a) She reported even HER PRIVATE INCOME. (156b) She reported even THERE HAVING BEEN A BROWN-OUT. (156c) She reported even THAT SHE COULD HAVE BEEN BEING

GRIPPED BY KONG.

-~""':~;;;~~~'~-"'';

'v

384 ROSS AND COOPER

5,

-t-. "i 1 .. 'M 'OM"" .... ')'

52

THAT 5HE COULD HAVE BEEN BEING GRIPPED BY KONG FIG, 11,17. Example(157)

We will assume that in these cases, the even is Chomsky-adjoined to the direct object NP of reported, as in (157), Fig. 11.17.

Like the adjectives in (132); report allows deletion of that.

(158) She reported (that) we were being followed by a sasquatch.

The question is, what would happen if the rule of "That"- Deletion wereto delete the that which introduces S2 in (I 57)? The result would be *(159).

(159) *She reported [even [ [SHE COULD HAVE BEEN NP 1 NP2 S2

BEING GRIPPED BY KONG] ] ] S2 NP2 NP 1

It might be thought that this could besaved by repositioning the even to the right of she. However, although the result, (160), is grammatical, it does not seem to us to have the same meaning as (I 56c); i.e., although (156c) can be used as an answer to (161), (160) cannot, in the same sense.

(160) She reported she even could have been being gripped by Kong. (161) Did she report anything else?

It would seem that there are only two ways of avoiding *(159): The first would be to make the rule of Intensifier Climb obligatory here, and the second would be to block the rule of "That" Deletion, if there is a preceding evr». '1 he first "solution" has nothing to recommend it. The second, although it initiall. sounds dreadfully ad hoc, may be on the right track, for note that it is oltcn impossible to delete that unless the thaI-clause immediately follows the predicate of which it is the complement [cf, (162) and (163) below].

11, LIKE SYNTAX 3l:io

( I62a) It was reported (that) they had found a solution.

in The Times}(162b) It wa.s reported { b Harold *(that) they had found a solution. y

(163a) I believe (that) we will win from the bottom of my heart. (163b) I believe from the bottom of my heart *(that) we will win. ( I 64a) I believe (that) we will win, and you believe (that) we will

lose. Gapping ~

(1Mb) I believe (that) we will win, and you *(that) we will lose. (165) What I believe is *(that) Horton may have a point.

So although there are some cases of constituents that can intervene between a verb and a deleting that [cf. the underlined elements of'( 166)], the asterisked sentences in (162) through (165) provide some independent motivation for an adjacency-to-matrix-predicate condition that might be appealed to in order to prevent (157) from becoming (159).

(I 66a) He told us (that) we were fools. (l66b) It seemed to us (that) you were wrong. (166c) I figured out ?(that) my phone was bugged.

Note, however, that if that cannot delete after even in (157), it will also be blocked from being deleted in (152), at least by the simplest statement of the rule. Thus, our earlier assertion that (152) can be converted into (155a), and later, by "Even" In, to (l55b), turns out, upon inspection, to require qualification.

What (152) shows is that even can be attached to the node S. The fact that (152) is synonymous with (155b) will, thus, have to be accounted for bya rule that moves intensifiers to post-subject position, over an optional comple­mentizer. If the complementizer is present, the rule is optional; if nothing intervenes between the intensifier and the subject, the rule is obligatory. The rule that has this effect is formulated in (167).

(167) Intensifier III

iUSI I {Ihat}X-·[ 0111.1' -[( )-NP-VY]]-Z )Seven S for S S also OPT 2 3 4 5 6 ~

I 0 3 4+2 5 6 Condition: If Term 3 is null. the rule is obligatory for all intensifiers except also, for which it ispreferred, and like. for which it blocks (in our idiolects).

~l~~~~=~~ij~n:ht~~;;,;;.,t;'. _.

iii

386 ROSS AND COOPER 11. LIKE SYNTAX 387

5, (171) "'John gave even his daughter a new BICYCLE. [in Jackendoff's (6.91»).r.

Jus' 52

( even ~~ ~ also } NP X - {'hal

for FIG.11.1B. Example (168)

It appears necessary to require that intensifiers only move inward when in such structures as (168), see Fig. 11.18, (i.e., when they are immediately dominated by an S-node that immediately dominates another S-node), for otherwise, even would be able to move after she in (157), producing a sentence, (169), that is grammatical, but which, like (160) above, does not have the same meaning as (156c).

(169) She reported that she evencould have been beinggripped byKong.

Let us recapitulate the points wehavemade, for they are complex. Wehave argued that all intensifiers must be allowed to be Chomsky-adjoined to S, on the basis of sentences such as (152), and that all intensifiers except /ike must subsequently be allowed to be moved to the post-subject position, by rule

11/ (167).Thus, we have rejected "Like" Out, and adopted (essentially)"Even "In (seesubsection Pauseless"Even ') concluding, therefore, that the behavior of f' like is more indicativeof the basestructure ofsentences than isthe behaviorof

I the other intensifiers. And like suggests that sentences must be ternary i ~

branching at the highest node, not binary (thus, we reject (146) in favor of !. (147», and right-branching thereafter. We believe the structure of the

maximally complex auxiliary, therefore, to be as it is diagrammed in (149).

i Previous Analyses of Even

I Jackendoff's Analysis. We now treat two of the previously proposed !

accounts of the distribution of even bearing most directly on the issueof the structure of the auxiliary.

Jackendoff (J 972) proposed the following condition (p. 249):

(170) If even is directly dominated by a node X, X and all nodes dominates by X are in the range of even. [= Jackendoffs (6.93)J

Jackendoff cited the following type of case in support of this principle. In (J 71), the noun BIC YCLE is not in the range of even, so the sentence is ungrammatical.

Jackendoff observed that if even is dominated by the NP that dominates his daughter, his principle, (170), accounts for this ungramamticality, for even would not C-command BICYCLE.

We find one difficulty with Jackendoffs proposal, namely, ifeven werenot attached to his daughter in (171), but instead were directly dominated by VP, as in (172), see Fig. I 1.19, then sentence (171) would incorrectly be ruled acceptable. The question is, can structures such as (172), in which even is a daughter of VP, be excluded in principle?

s

VP

HP v NP NP

I I DDFIG.11.19. Example (172) John QOvo even hi. dauQh'or a no.. BICYCLE

Jackendoff (1972) was not very explicit about where even is to be introduced in remote structure; the only comment he made on this topic was:

(173) These words [i.e., even, only, and just-JRR&WEC] can occur before NPs and in the auxiliary. (p. 247).

However, one of the sentences that he cited allows us to infer that there must be some evens that are immediately dominated by VP. The sentence. Jackendoffs (6.97), is repeated in (174);

(174) John will have even given his DAUGHTER a new bicycle.

Jackendoff remarked in connection with (J 74): "The followingexamples bear this out: even after the second auxiliary is unacceptable with subject focus, although other foci are possible [po 251)."

We agree with Jackendoff (1972) that even is linked to DA UGH7ER in (174), but note that, although it might betrue for some theories of the English auxiliary (e.g., in the classicalanalysis of Chomsky's Syntactic Structures, in which the node Aux is expanded as in (175),

(175) Aux - C (M) (have-en) (be-ing) (be-en)

,,"~~"

3<." ROSS AND COOPER

that even in (174) is occurring "in the auxiliary'), this is not the case for the analysis of the auxiliary that Jackendoff adopted in his chapter on adverbs. There, on p. 76, he proposed the following rules:

(I 76a) S - NP - Aux - VP [Jackendoff's (3.122)] (I 76b) Aux - Tense - (Modal) [Jackendoff's (3.123)] (I 76c) VP - (have - en) (be - ing) V - (NP) ... [Jackendoff's (3.124)]

There is one more rule that affects the structure of the auxiliary: J ackendoff's rule of"Have- Be" Raising [Jackendoff's (3.127)], which has the effect of putting either have or be under the node Aux, if that node does not contain a modal and consists entirely of Tense.

Because there is a modal (will) in (174), however, "Have-Be" Raising will not be relevant, and the remote structure of (174) would have to be as shown in (177), Fig. I 1.20.

The question now arises as to how (173) is to be amended in such a way as to allow (177). If we simply add the words "or in the VP" to the end of (177), we will provide an account for (177), but we will also incorrectly allow *(171) to be generated, as well as a sentence such as *(178a), assuming that it has a structure such as that in (178b), Fig. 11.21.

S

NP Au! /\ VP

Ter "'T' L

/ r-; NP

-, ~~16

John Pres Win hove even en 9,ve h,. DAUGHTER a ne'llf bicycle

FIG. 11.20. Example (177)

s

I NP vP ··~t

~\ PP PP ~~ A <; .W

t;. P NP A (2\ ':zJ,~

j..J. -_.> /:I~-'f h. troded NUTS to the inhabitant. "ve" Ice foot boll ticket'S In NO'lernb,'~t;

FIG. 11.21. Example (178b)~t' ~ h ;~; ~~.

U

11. LIKE SYNTAX 3...

(l78a) "'He traded NUTS to the inhabitants even for football tickets in November.

The alternative would seem to be to replace (173) by a very unsatisfying disjunctive list of categories, such as that in (179):

(179) The words even. only and just can occur either before [i.e., as left daughters of?] NP or before [i.e., as left sisters of]:

-en [cf. (177)] -ing'? V [cf. (95b) above] VP [for sentences such as "Jim may even have been

eating GRAPES'l Aux [for sentences such as "Jim even may have been

eating GRAPES'l Adj [for sentences such as "Tex may be even ANGRY at

me"] Adv [cf. (93)] P [cf. (94b)] PP [cf. (75e)] Complementizer [cf. (152)]

Possibly some other less disjunctive alternative can be devised that stays within the spirit of Jackendoff's (1972) analysis. We will not search for one here, but rather sketch an approach to the distribution of even that seems to be more fruitful to us.

Our basic disagreement with Jackendoff (1972) lies in the fact that such statements as (173), or the more exhaustive (179), are too close to the surface. We propose. to replace these with a characterization such as that in (180), which is essentially semantic.

(180) Even can modify any semantic predicate or argument.

This characterization will have the effect of allowing even in remote structure to modify any lexical category, namely. nouns, verbs, adjectives. or (meaningful) prepositions, as well as NPs. But it will lead to some surface problems: although (181) is a possible remote structure (because window is a noun), it is a bad surface structure.

(181) "'They may fingerprint Ben's even WINDOW.

"This would he necessary in order10 generate scntcuccs such as (i). whose remote stuH:tun.'s

are parallel to (177) in all relevant respects: (i) "John may be evengivinghis DAUGIHE R a new bicycle."

..~.:.

• -'''1:: ',:~~f·~>';-.'· . ~~:T-I'"

390 ROSS AND COOPER

We propose to make the rule of Intensifier Climb obligatory in such cases: Even must be adjoined either to the, direct object NPofjlngerprint or to the constituent following may in (18\), which we would analyze as a clausal remnant, following Ross (1969), not as .a VP. The result would be either (182a) or (182b).20

(I 82a) They may fingerprint even Ben's WINDOW. (182b) They may even fingerprint Ben's WINDOW.

That is, in our view, the surface positions of even are most effectively accounted for byadopting a highlyrestrictedsemantic characterization of the remote elements that it may modify and then by letting the rule of Intensifier Climb (and the subequent rule of Intensifier In) produce surface results that may not reflect the underlying semantic regularities in any direct way.

The most salient problem for this analysis of even is the fact, noted by Anderson (1972), that even can be associated with surface structure constituents that are not elements of semantic representation. An example is (183).

(183) Our new boss is a dream; he's pleasant, he doesn't make you take shorthand, he gives half-hour coffee breaks, and he's even easy to get a raise out of, if you wear short enough skirts. [Anderson's (7)]

Under the "Tough" Movement analysis, which we assume to be correct (but any rule making derived subjects would have the same force), easy to get a raise out a/corresponds to no piece of semantic representation, as Anderson pointed out.

One possible solution within the framework advocated here is to say that the relevant sentence in (183) has a history like that shown in (184).

(184a) Approximate remote structure: even (for one to get a raise out of him) is easy for one ~ (via Equi, Deletion of/or: one after easy and "Tough" Movement) Intensifier Climb

(184b) He is easy even to get a raise out of. ~

(184c) He is even easy to get a raise out of.

It is extremely unclear to us at present exactly how the difference ill meaning between(184a) and (184c)(which issubtle but real. we fcel), is to he represented in a formal semantic representation. The latter sentenceseems to predicate something of he. whereas the former one does not. But how is

lONote that Intensifier Climb must be made obligatory in other contexts as well: 'neitherall" nor just can comfortably precedean NP that is the object of a preposition [cf.(96c)and (114a\ilJ

11, LIKE SYNTAX 391

"predication" reflected structurally? If there were some extra semantic primitive, presumably an empty predicate, in the semantic representation of (184c) [something that is not in the representation of (184a)], then possibly even could, within the confines of (180). be said to modify this element. Or possibly even modifies the complement of easy underlyingly.as suggested by the representation in (184a), with the rule of Intensifier Climb being globally constrained not to change (184b) into (l84c) except in the presence of whatever semantic structure is relevant to producing this predication interpretation. Our present understanding of this area is too limited.

Let us recapitulate our remarks on Jackendoffs (1972)analysis. We have argued that Jackendoff's statement (173) is inadequate to account for the location of even in (174) and that there is no simple way to modify (173) in such a way as to account for (174) without also generating (171). And any such list as (179) should be avoided if at all possible. Our suggestion is to adopt the right-branching analysis ofthe auxiliary that is implicit in(149)and to allow the rule of Intensifier Climb to apply to position intensifiersto the left of any auxiliary except the first.

Anderson sAnalysis. The basic condition on even that Anderson (1972) stated is cited in (185):

(185) We could accordingly restrict the scope of even to constituents that are either directly adjacent to even, or dominated by a node that is [p, 899].

Using the notion of C-command introduced by Reinhart (1976), we can recast Jackendoff's (1972) definition [(170)J and Anderson's (1972) as in (186a) and (186b), respectively, which will more clearly bring out their differences.

(186a) Jackendoff: even C-commands any focusedelement with which it is associated.

(186b) Anderson: even C-commands any focusedelementwith which it is associated, and it must be adjacent to either thc focused element. or to some node that dominates the focused element.

Why did Jackendoff. whose work followed Anderson's. adopt a less restrictive condition on even? The reason. we believe, isthat Anderson did not consider sentences such as (187).

(187) JO HN willeven have given his daughter a new bicycle. [~ Jacken­doff (6.96)J

This sentence is clear evidence against Anderson's adjacencyclause in( 186b).

~'F :_iiDtIilIJ!!ti!iMi¥!!f!ii!W~4.;*Pli;;;,*M';;N.®''''

392 ROSS AND COOPER

On the other hand, such sentences as *(171), "*John gaveeven hisdaughter a new BICYCLE," which are equivalent in all respects to sentences such as (188), are what impelled Anderson (/972) to argue for the adjacency clause. As we have seen, Jackendoff (1972) can apparently only avoid generating sentences such as (171) and (188) by assuming that the evens in them are attached to the following NP. However, sentences such as (174) suggest the need for postulating YP-dominated evens in Jackendoff's system, and an unwieldy disjunctive specification soon heaves into view.

(188) *J ones eats even Skrunkies for BREAKFAST. [Anderson's (/3c)]

The way we propose to resolve this conflict is basically in keeping with Anderson's (1972) condition, for notice that any intensifier that has undergone Intensifier Climb will be in an output configuration that will satisfy Anderson's condition. Because the stress facts with also provide independent support for Intensifier Climb, we assume that sentence (187) must be accounted for by another rule, because it cannot be produced by Intensifier Climb from (189), which we assume to Uriderlie it.

(189) [Even JOHN) will have given his daughter a new bicycle. NP

If Intensifier Hop applies to (189),(I 90a) willresult, and the only place that even could climb to is the S node, which would produce (190b).

(190a) [JOHN, even,] will have given his daughter a new bicycle. NP

(190b) JOHN will have given his daughter a new bicycle, even.

Alternatively, wecould apply Intensifier Climb directly to (189), producing (19Ia), to which Intensifier In would obligatorily apply to produce (191 bo).

(19Ia) [Even [John will have given his daughter a new bicycle] ] S S SS

(19Ib) JOHN even will have given his daughter a new bicycle.

But no rules yet given can produce the desired (187). Before we turn to proposing a rule that will, let us take issue with

Jackendoff ( 1972) on a grammaticality judgment. J ackendoff asterisked ( IR7) if the el'e/1 follows have. He remarked:

(192) This analysis provides confirmation of the.theory of the auxilinry presented in section 3.8, in which it was argued that, in the surface

11. LIKE SYNTAX 393

structure, the first auxiliary is dominated by S and the other auxiliaries are dominated by YP. This analysis predicts that the subject is within the range of even only if even comes before the second auxiliary, since otherwise even would be dominated by YP. not S [pp. 250-251].

However, although we can agree with Jackendoff (1972) in finding even after have in (187) slightly less acceptable than before have [cf. (193»). we would not rate it fully ungrammatical, as he does. Its awkwardness seems to us notto have to do with even so much as with will; with some other modals, as in (194), the sentence is unexceptionable, to Our ears.

(193) ?JOHN will have even give his daughter a new bicycle.

should (?n't)j

(194) JO HN Icould (n't) have even given hisdaughter a new bicycle. may might (n't)

Other examples confirm this intuition:

(195a) Were OTHER STUDENTS eligible for the grant? (l95b) Yeah, SOPHOMORES could have even applied for it. (196a) Could only TEX have gotten to see the princess? (196b) No, SULLY might have even been able to do it.

It seems; then, as if the full set of facts pertaining to evens that follow a focused subject are in conflict with even the weaker condition imposed by Jackendoff (1972), for even in (194) through (196) will not C-command the subjects, either under Jackendoff's analysis of the auxiliary or under our right-branching account. Thus, some new rule is needed.

The rule must move not only even, but also also: cf. (197).

(197a) Could anyone else be indicted? (197b) Yeah-also PROFESSOR BEEBE could be indicted. (197c) Yeah-PROFESSOR BEEBE could also be indicted.

However, ()1l~J! is not affected by the rule [thus ( J98a) cannot become(198 bj], and neither is like. for us usually [cf. the previous discussion of(22) and (23)J. Thus, the rule is highly ad hoc: we formulate it as in (J 99).

(198a) Only HAROLD can get us out of this. (198b) *HAROLD can only get us out of this.

_~ ~~~~~~::~~~~~:.'.?2~.~~.

394 ROSS AND COOPER

(199) X[bFv:~fille: ;:lr :(['~S:ux]_ {also J + Tense

NP OPT I 2 3 4 => o 2 3+1 4

~: , Note that this rule produces structures such as (187), which would be ·1. forbidden by Anderson's (1972) condition, though not by JackendoffsI,

I

(1972). However, as it stands, it will not generate (194) through (196), which would also be blocked by Jackendoffs condition. Thus, it will have to be amended in some way to produce these sentences. As Jackendoff noted, even and also are barred from being three auxiliaries away from a focused subject . (to say nothing of four):

(200a) Could anyone else have been being shadowed?

even} .(200b) *Yeah-JESSICA could have been also being shadowed. {

. {even}(200c) **Yeah-JESSICA could have been being also shadowed. ,i

lIlt

n It appears, however, that what is involved here is not just a simple counting of auxiliaries, for note that even and also refuse to follow the bare stem be, or the gerund being [cf. (20 I) and (202)].

(20Ia) Could anyone else be being shadowed?

even} .(20 Jb) *Yeah-JESSICA could be also being shadowed.{

(202a) Is anyone else being shadowed?

(202b) **Yeah-JESSICA is being{evten} shadowed. II~. . . a so

I"

In fact, they do not even like it very much after been:

working hard in that room

(203a) Has anyone else been ?sick a nuisance

I~.•

11. LIKE SYNTAX 395

??working hard

even} *in that room (203b) Yeah-JESSICA has been also{ *sick

*a nuisance

Rather than "ad hoc"-ening rule (199) even more, by inserting an optional (have) at the end of Term 3 (which would generate (193) through (196), but would predict that (203b) is totally bad, even with working hard, which it is not), we will briefly cite some parallel facts involving quantifiers, which suggest to us that the solution to the problem of what auxiliaries even can follow lies elsewhere than in making amendments to (199).

There is a process known as Quantifier Floating, which has the effect of converting (204a) into (204b).

(204a) {~~Ith} of them are snoring.

(204b) They {~~th} are snoring.

We would agree with Postal's (1974b) assessment of the way this is effected: Quantifier Floating works by ascending the object NP of alto make this NP the derived subject of are snoring. 21

We also find all and both after are, and, in fact, after any tensed auxiliary

verb:

a ll ', . {(205) They are both]. snormg.

all } b .(206a) They could { both e snoring.

a ll \ {taken Benzedrine } (206b) They have { both] *work to get done-? .

h ,{all}(206c) T ey don '. both. snore.

If the second auxiliary is have, we find floated all and both to its right:

"Thc fairly extensive treatment of Quantifier Floating by Postal (1974b) is challenged h~

Ficngo and Lasnik (1976). Sec also Maling (1976) for some important related points. 22Th is is bad because when workto hedone follows have.haveis a main verb, not an auxiliary.

~.

1;~j~\~7··

396 . ,055 AND COOPER

can't must

all }(207) They {should have both been lucky. { way etc.

However, if the second auxiliary is the bare stem be, or the gerund being, the quantifiers cannot follow it:

will must

(208a) ·They {should be {allboth} spies,23 may etc.

(208b) "They are being {~~th} pests.

With quantifiers, the position after been is bad, even if a progressive follows:

working hard in that roomall } .24(209) "They have been both{ Turkish nuisances

llFor some reason. after call" he and couldn't he. things are often looser than with other medals. with all being freer to Iloat to the right of be than is both. Thus, for us. the following contrasts obtain:

?ean't ~ (i) They ?couldn't be all spies.

?'eould ~ ?'! ea n 't ~

(ii) They 7?couldn't be both spies. 'could ~

We have no explanation for this differential behavior of call and could.

l4For reasons we do not understand. however. some ofthe sentences in(209)sound bcucr w i'h medals and all:

11lll ' ! I I working hard

. . could in that room (I) 1hey I I j have been all .,,',' kisl , .

S iou ( .. ur IS 1

1may "uuisanccs

IllUSI "wurking hard I \(ii) Th (ould I h h I '/'Iin that roo III

U cy should 'ave cen at, '!'!Turkish

may *nuisances

These facts leave us with mouths agape.

11. LIKE SYNTAX

The parallels between (20I) through (203) and (207) through (209) appear to us to be more important than the differences. Accordingly, wewould like to give intensifiers and quantifiers highly similar analyses. In the case of quantifiers, Postal (1974a) has argued that Quantifier Floating is a cyclic rule. The motivation for this claim arises from variants of sentences such as (210).

(210) It is .Iikely that it will appear to you that {~~th} of them are snormg,

Assuming that this sentence is underlain by a remote structure along the lines of (211), see Fig. 11.22, we note that since both appear and (be) likely are Subject Raising (A-Raising) tr iggers.s! NP 3 can be raised once, to become the superficial subject of appear, as in (212), or twice to become the superficial subject of (be) likely, as in (213):

SI

~ NIP is likely

52

~ NIP will appear 10 you

S,

~ NP3 ore snorinQ

L all of NP4

IFIG.11.22. Example(21ll them

(212) {all I' I beIt ..IS likely t hat both } 0 t iern WI'11 appear to you to

snoring.

(213) {~~th} of them arc likely to appear to you to be snoring.

l'For an extensive t!eatlllcnt of the process of A·U"i.<iIlK, ef. Postal (19743),

.<.}~~r!;~f:'

398 ROSS AND COOPER

Now if weassume that Quantifier Floating, whichconverts(204a)to (204b) by ascending them to become the superficial subject of (be) snoring, is an optional cyclic process, then we would expect to find that all and both could show up before (be) snoring (whose deep subject they werepart of) or appear [because the quantified NP can be raised once, to becomeappear's subject,as in (212)],or before (be) likely [because the quantified NP can be raisedagain, as in (213»). All of these predictions are in fact attested: cf. (214).

(214a) It is likely,that it will appear to you that they all are snoring. (214b) It is likely that they all will appear to you to be snoring. (214c) They all are likely to appear to you to be snoring.

In fact, although the quantified NP has been raised twice in (214c),there is no necessarycorrelation betweenthe number of applicatons of A-Raising and the position of all in surface structure. If Quantifier Floating applies on the first cycle,with they subsequently being raised twice,(215a)will ensue, and if all of them is raised once before the quantifier is floated, with they subsequently being raised once more, (215b) will result.

(215a) They are likely to appear all to be snoring. (215b) They are likely all to appear to be snoring,

Postal (1974a) notes (p. 158, Footnote 18) that, given an analysis of auxiliaries in which each auxiliary is a main verb that undergoes A-Raising. the cyclicalQuantifier Floating analysis would predict that quantifiers should be able to precede any auxiliary. The sentences he cited in (i) of Footnote 18 are reproduced in (216).

(216a) A]] of the rockets will have been decaying for some time. (216b) The rockets will have been(probably) all decaying for some time, (216c) The rockets will have all been decaying for some time. (216d) The rockets will all have been decaying for some time. (216e) The rockets all will have been decaying for some time.

He observed: "(ib) [= (216b)] without probably is bad for many speakers because of an output condition blocking intensed be directly before a Q, that is:

*be ({ ~~g}) Q"

We find ourselves in general agreement with the proposed output condition, with the exception that the three conditions abbreviated in Postal's

11. LIKE SYNTAX 399

condition are different in strength. No one we know of tolerates quantifiers after being, with the bare stem be being only slightly less impossible, perhaps [cf. (208»). The situation with been is non-uniform: Compare (209) with the sentences in Footnote 24.

Although we know of nothing that would completely explain these differencesin strength, the followingobservations may be of some relevance. The passive auxiliary be allows its complement, the passive participle, to be deleted by VP Deletion, except after being.

(217a) John was arrested, but Christina was not (arrested). (217b) John has been arrested, but Christina has not been (arrested). (217c) John will be arrested but Christina will not be (arrested). (217d) John is being arrested, but Christina is not being *(arrested).

The same paradigm obtains for predicate nouns and adjectives:

POlite} . . {POlite}(218a) John was t • but Christina was not ( t ).{ apes apes'

POlite} . . { polite} (218b) John has been t .but Christina has not been( ).{a pes a pest

(218c) John will be{polite but Christina will not be ( {POlitea pes t}, a pes t}).

(218d) John IS emg{POlite}t' but nsuna IS not bemg*( {POlite} ).. bei Christi . a pes a pest

This does not seem to be a property only of being when it follows the progressive auxiliary: In (219), we find parallel facts when the -ing on be is induced by the matrix predicate continue.

(219) Tim was polite at first, but he has not continued {*tbo .be} .el1lg

We note that the tight bond between being and its object cannot be sundered by ripping rules, either. Note the sentential which-clauses related to the sentences in (217).

(220a) John was arrested, which Christina was not. (220b) John has been arrested, which Christina has not been. (220c) John will be arrested, which Christina will not be. (220d) *John is being arrested, which Christina is not being.

We might propose the following positive output condition:

~5'~'=~ ..,

40C 5S AND COOPER

(221) The Great Chain of "Being" If an instance of the verb he appears with the suffix -ing; whatever immediately follows this he in remote structure must immediately follow it in surface structure.

This constraint will not only block deletions such as those in (217) through (219), and choppings, such as that in (220), from taking place but also such "insertions" as **(200c), *(202b), and **(208b).26

In addition, the constraint may shed new light on the tag question construction, which, despite a 20-year history of investigation within generative grammar, still presents many mysteries [cf. (222)].

(222a) Hank might have been being bluffed, mightn't he? , (222b) Hank might have been being bluffed, mightn't he have? (222c) ?Hank might have been being bluffed, mightn't he have been? (222d) *Hank might have been being bluffed, mightn't he have been

being?

It is too early to tell, but it may well be the case that there are conditions on be and been which are similar to (221), but weaker. If there is in addition an interacting hierarchy of rule strengths, with deletion rules, such as VP Deletion being "stronger" (i.e., able to apply in more environments) than chopping rules, such as the one that forms the which-clauses of (220), and with chopping rules being in general stronger than insertions, then such differences in acceptability as (217c) or (220c), on the one hand, and (208a), on the other, may find an explanation. The situation, thus, would parallel the general state of affairs in island constraints; in both areas, nondiscrete treatments would be necessary.

However, to pursue these matters further here would take us too far from the task at hand, namely, accounting for the positions in which even and also can appear in the auxiliary. What makes floating quantifiers relevant to the distribution of even and also are the parallels between(193) through (196)and (207), or between *(20I b) and *(208a), or between *(202b) and **(208h) These sentences show that even and also can appear roughly where all and both can.

We must say "roughly," because the parallelism is not perfect: Quantifiers seem to be able to follow been slightly better than do intensifiers. Compare

:!"\Vc have placed quotation marks around insertions. because, in line with our acceptance III Perlmutter and Po~tal's (197H) ascension analysis of floated quantifiers, constraints on surlace order of auxiliaries and quantifiers will cit her 'have to he expressed as output conditions, a, Postal did. Of as constraints Oil Raisin); (0 SIl'~jC('1 l'osuion, For the present.this point of detail i... irrelevant. and "'0 continue to speak of "inserting" quantifiers alter auxiliaries and other verr«

11. LIKE SYNTAX Ii

(200b) and the sentences in Footnote 24 above. Furthermore, when we examine the behavior of even and also in nests of Subject-Raising constructions, like (211), we find that quantifiers can be separated bya longer string of Subject-Raising triggers from the subject NP they quantify over than can even and also be separated from a focused subject with which they are associated. Thus, although (223a) and (224a).can be answered with (223b)and (224b), respectively, for all speakers, and by (223c) and (224c) in many dialects, we know of no speakers who would allow (223d) or (224d).

(223a) Is EVERYBODY likely to appear to me to be snoring?

even is t.(223b) Yeah-PRESIDENT HUGHES . likely to appear to IS even ~

you to be snoring.

(223c) %Yeah-PRESIDENT HUGHES is likely ~even to ~ appear 1to even \

to you to be snoring.

(223d) *Yea~~ve~~~rSIDENT. HUGHES is likely to appear to

you be snoring. to even

(224a) Is only MRS HUGHES likely to appear to me to be snoring?

(224b) No, PRESIDENT HUGHES ~~Isol is ~ likely to appear to /IS a so \

you to be snoring.

(224c) %No, PRESIDENT HUGHES is likely j alsolto ~ appear to ito a so \ you to be snoring.

(224d) *~~~'oPt~E~5bIDEN'.r HUGHES is likely to appear to you

e snoring. to aIso

The (d) versions of the sentences here are to be compared with (215a) and the (c) versions with (215b).

There are, then, two cases that seem to show that the general parallelism between quantifiers and intensifiers is overlaid by a tendency for quantifiers to occur in more environments than intensifiers.

Let us now turn to a detailed examination of (223c) and (224c), for these sentences have an important bearing on the analysis of evl'll and also: they show that neither Jackendof]'s (1972) nor Anderson:~ (1972) condition Oil

"even" call he true of surface structure. Rather, the range of even (and also)

11. LIKE SYNTAX 403402 ROSS AND COOPER

seems to have to be determined cyclically. We propose the derivation for (223c) shown in Figs. 11.23 through 11.26.

The only problem that is posed by our cyclicalanalysisof even and also is the ungrarnmaticality of (223d) and (224d).The first of these could arise, for instance, if the optional rule of Intensifer Climb were allowed to apply on the S3 cycleof (225a).lntensifer In, an obligatory rule, would then have to apply, and even would remain as a constituent of S3.

At present, the only remedy we can suggest is a global filter, which would mark as ungrammatical any surface structure containing a focused surface subject and an even or also that is more than one clause down from this subject, if the main verb of the clause in question is not an auxiliary verb.

This is an extremely ad hoc restriction, but the (c) versions of (223)and (224) seem to force a cyclical treatment, and we know of no alternative

(225a) Approximate remote structure

5,

~ NP,. is likely

I S2

~ T

appear 10 you

ROISlng

A ~

S3

(NP3 becomes the

~ subject of appeor 10 you 27 J

be sno,inQ

even NP4

~ PRESIDENT HUGHES

FIG. 1'1.23. Example (225a)

"When the subject of the complement of appear ascends to become the derived subject of appear, the remnant of the old co~plement (to be snoring) becomes a chomeur, which we indicate in (225b) by the notation NP,. See Perlmutter and Postal (1978) for more details.

A is likely Jnlensifler Climb ond

T Intensifier In 28

---~ S2

{eyeo teo\/e5 NP4 and moves 10 the left of appeo r ]~NP2l\ .~....,ou

/

1 even NP. 53

~ 6 to be snor inQ PRESIDENT HUGHES

FIG. 11.24. Example (225b)

5,

~ Raising

Nfl is likely ~

[NP 4 becomes the 52 subjecl at ~J

~ NP. even appear to you NP12

~ 653 PRESIDENT HUGHES

to be snorinQ

FIG. 11.25. Example (225c)

2'1t might be thought that it is possible to achieve the effectof the two rulesof lntcnsificr Climb and Intensifier In by allowing NP, to undergo the type of NP-asccnsion that underlies Quanti/iff F1oating.This would convert even into a chomcur, and it (and 'also) could be assigned the same position in the linear ordering as all and both. However, we arc dubious. On the one hand. Intensifier Climb is needed in objects. independently of what happens in subject position: Sec the sentences in (74). (75). (97), (IUa). and (114a). Furthermore. on the basis of sentences such as (152). we have argued that e"en must be able to beclimbed to attach to higherS nodes. This being the case. Intensifier In will be necessary to repair sentences such as (150). Thus. we sec no advantage to postulating that the conversion of (225b) into (225c) is accomplished by Quantifier Floating.

,~_:,~':J~

r·-·.../' .,

'. ~

11. LIKE SYNTAX 404 is AND COOPER

5, 5,

6 ;'''~'' r PRESIDENT HUGHES

~ /'.

even 10 appear 10 you NPz

I 6

S3

10 be snorlnv

FIG. 11.26. Example (225d)

account in which there is no equally ad hoc condition that would correspond to the global filter that we have adumbrated above.

Some sugar coating for the bitter pill of the global filter is available, however: The cyclical analysis of even and also obviates the need for the rule of Intensifier Post Tense [( 199)]. Recall that this rule was postulated in order to account for sentences such as (187), "JOHN will even have given his daughter a new bicycle," because the even here violates Anderson's (1972) adjacency clause, which otherwise seems desirable. But if auxiliaries are treated as main verbs, then this sentence will derive from a remote structure such as (226), Fig. 11.27.

This cyclical analysis of nonadjacent SUbject-linked intensifiers thus kills two birds with one stone: No rule (199) is necessary at all, let alone an ad hoc optional have in Term 3 of that rule. Both sentences such as (187), which violates Anderson's adjacency condition, and sentences such as (193) and (194) through (196), which even violate Jackendoff's (1972) weaker C­command condition, drop out of the analysis as a consequence of a main-vern analysis of auxiliaries, coupled with the principle of the transformational cycle.

Further Parallels Between Intensifiers and Quantifiers

Before concluding our discussion of the English auxiliary, which has been based in part on some similarities between intensifiers and floated quantifiers.

1\ NP V

II I

I S2 will

NP/\

V

) \ 53 hove

~l\ .;~'"' ", '00,"", • ... ""."

even T FIG. 11.27. Example (226) JOHN

let us draw attention to an additional point of similarity between these two classes of words.i?

As we have seen above, Intensifier Climb can move only from its original location in (204), adjacent to PLA YBOY, to Chomsky-adjoin it to any higher 'constituent.

(227) They could [have [read only PLA YBOY]].

Thus, Intensifier Climb can by itself produce either (228a) or (228b) from (227), and if only is first climbed to be adjoined to the highest S node and subsequently moved to the post-subject position by Intensifier In. (229) will

result.

(228a) They could [have [only [read PLA YBOy]]]' (228b) They could [only [have [read ?LA Y80 YJ]J.

(229) They only could have read PLA Y80 Y.

"We assume here that thc rule that has the effect of moving ("'/'11 and also rightward II'''''' focused subjects can be identified with Quontifur Floating. despite thc fuct that quantifiers ;11,'

more flexible than intensifiers. as we have noted. Nothing in the discussion rests on t hi­

assumption. however.

;:X!-:-;1':(1{'':1;''~~:;!ilt{.~,~,·\

.~"~'..

406 ROSS AND COOPER 11. LIKE SYNTAX 40"7

Thus, in a sentence with an object-linked intensifier. there are usually three positions to the left of that object in which the intensifier can appear.

Let us now consider a sentence with a subject-linked intensifier, such as (230).

(230) Even THE NEPTUNIANS could have read it.

This sentence derives from a structure similar to that in (226); If Quantifier Floating applies on the lowest cycle,(231 a) will be generated, and if it applies on the next lowest, (231 b) will result.

(23Ia) %THE NEPTUNIANS could have even read itJO

(231 b) THE NEPTUNIANS could even have read it.

And if Intensifier Climb applies to the remote structure of (230),Intensifier In will obligatorily produce (232).

(232) THE NEPTUNIANS even could have read it.

Thus, we see that there are usually also three places to the right of a focused subject in which a subject-linked intensifier can appear.

Consider the situation in which there are two intensifiers, one associated with the subject and one with the object. The restriction operative in such casesseems to be as follows; If weimaginecrepe-paper bands running from an intensifier to the focused NP with which it isassociated, then such bands may not cross. To see this crossing restriction at work, imagine the possible responses to (233a), all of which are variants of the answer in (233b).

(233a) Could ANY OTHER OUTWORLDERS have been reading only PLA YBOY? .

(233b) Yeah-even THE NEPTUNIANS could have been reading only PLA YBOY. '

If even remains to the left of THENEPTUNIA NS. ani)' can appear in the normal three places, as is indicated by the three lines from its prenominal location to the carets in (234). Fig. 11.28.

If even has undergone Intensifier Climb and Intensifier In, then there are only two possible locations for only (Fig. 11.29).

Note that the variant of (235) in which only precedes e1'en could be excluded by the proposed constraint, for the crepe-paper bands from TilE

"This type of sentence is untarnished for us. hut we have prefixed it with a percentage sign to indicate our belief that this is a controversial judgment.

Yeah - Even THE PLAYBOY.""""..,~'"

FIG. 11.28. Example (234)

Yeah - THE PLAYBOY.

"""~,.

FIG. 11.29. Example (235)

PLAYBOY.Yeah· THE II eouldII even/lhave /I,eod NEPTU~N1ANSonly II II

* II

FIG. 11.30. Example (236)

NEPTUNIANS to even and from PLA YBOY to only would have to cross, and such a crossing would be prohibited by the informally stated constraint given above. However, we are inclined to think that this sentence should instead be blocked by some form of "one-per-slot" constraint, because it is also impossible for only to immediately follow even in (235),despite the fact that no crossing bands can be found, Thus, in what follows, we point out the various' occurrences of the one-per-slot constraint as they arise, although we do not propose a formal constraint, because our investigation of its range isat present far too preliminary."

A clearer instance of a band-crossing violation can be seen in (236) (Fig. 11.30), where even has quantifier-floated on the second cycle, as in (231 b). Here, the reason that only cannot precede could, as it generallycan, is that the even-band and the only-bandJ2 cross at the ··ed circle.

Some final cases of a band-crossing violation can be seen in (237) (Fig. 11.31), where even has been floated on the lowest cycle, as in (231 a).

It is worth pointing out that it is not enough for the linear order of the intensifiers to be the opposite of that of the focused elements with which they are associated. This point can be seen on the basis of the sentences in (23Hh). which are versions of (233b) that would have arisen by applying Intensifier

.lINote. for instance. that whereas onlv cannot precede even in (2:16). it can follow it. yc: presumably both are in the "same sial."

·"Here we are considering the band from e"ento THE NEPTUNIANS and the band frlll11 only to PLA YBOY. respectively.

~~~~.

""".

11. LIKE SYNTAX 409408 RO::,;:, AND COOPER

" Veah - THE PLAVBOY. ",,"ccw.;:reodonly .. * * M ?.

M

* FIG. 11.31. Example (237)

Yech - THE NEPTUNIANS. even. could have read only PLAVBOV.

?Yeoh - THE NEPTUNIANS,,~IY PLAVBOV, even.

FIG. 11.32. Examples (238a) lop and (238b) bottom

Hop to (233b), producing (238a) (Fig. 11.32), to which Intensifier Climb would apply to adjoin the hopped even to the S node.

These sentences are certainly awkward, but they are not unintelligible. This fact shows that bands, or their functional equivalents, are necessary for any formulation of the restrictions on multiple intensifiers: One could not say anything like (239).

(239) If there are n foci in a sentence, and n intensifiers associated with these foci, the sentence will be ungrammatical if there is any difference in order between any two foci and their corresponding intensifiers.

The reason that (239) is inadequate is apparent. There are differences in order of the kind mentioned in (239) both in (236) and (237) and in (238), but only those of the former type produce violations. What distinguishes the two types is the notion of bands. and of crossing, both of which we conclude are essential.

We 'will now show how the band-crossing constraint interacts with the floating of all and both, Consider the sentences in (240).

(240a) Both (of) the Neptunians could have read only PLA YBOY. (240b) The Neptunians both could have read only PLA YBO Y, (240c) The Neptunians could both have read only PLA YBOY, (240d) The Neptunians could have both read only PLA YBOY.

These sentences show the widest set of environments for floated quantifiers'. IIowever, as was the case with a subject-linked {,I'en that was not adjacent to the subject. we find that if only undergoes Intensifier Climb, both cannot freely float to its right. this restriction is documented in sentences

The Neptun.ons bothA could l\,ovel\~Odh only PLAYBOY

I

~<:»:.. /",/ //

FIG.11.33. Examp!e(241)

PLAYBOY,., "".C'V'. FIG.11.34. Example (242) ,

The NePtU\iWnianS~COUld~hov.r.bOlh"reOdonly PLAYBOY.

;'* FIG. 11.35. Example (243)

?

(241) through (243), Fig. 11.33, 11.34,and 11.35,where the designated circles again denote band-crossing violations of the type encountered above,

What is immediately apparent here is that although quantifiers and intensifiers can produce band-crossing violations, they are much milder than those produced by two intensifiers.

One might be tempted to conclude that this isa transderivational fact; after all, although (244b) is ungrammatical when read as a band-crossing variant of (244a) (see Fig. 11.36), it is fine as a variant of (245),

(245) Only BILLY was writing even FRENCH poems.

And it could be pointed out that although in English, nonadjacent quantifiers can show up to the right of the NP they modify, they cannot show up to its left33 (see Fig. 11.37), Thus, it might be said, since both in (241)through (243) can only be linked to The Neptunians, no ambiguity can arise. so the sentences are all uniformly better than intensifier-crossed examples like(236) and (237), That is, it might be held that band-crossing violations are essentially a kind of no-ambiguity constraint."

.1·'There is no general prohibition against such a situation arising. In French, (i) can become (ii), via the rule of J.·7Ims (cf Kayne, 1975).

(i) II enri a voulu lire tous d'eux.

Henry has wanted to read all of them. (=wallted]

(ii) Henri u ious voulu les lire. Henry has all wanted them to read,

-"Sec Hankamer (1973) for discussion bearing on the avoidance of ambiguity.

;',I,,!~jJ.ii~~"t~~~f "-;('1"":""~lI"'?1~~fb_~~~~~e~;g1~~~~~~~~~1!i~!f.~~~§:~r:\~,~:'{::~"

410 ROSS AND COOPER 11. LIKE SYNTAX 411

Even BILLY wos writinQ onl~ FRENCH poems. that Anderson's (1972) more restrictive adjacency condition (I 86b) is to be poems. preferred to the less restrictive C-command condition proposed by

Jackendoff (1972) (I 86a), but that neither condition is adequate. if FIG. 11.36. Example (244a) top formulated in terms ofsurface structure. Rather, based on the grammaticality "WJ'" and (244b) bottom

"'~.. ,.. "","

FIG. 11.37. Example (246)

However, such a conclusion would be hasty. For if we recall that only cannot undergo Quantifier Floating [thus (247a) cannot become (247b)], we can see that there is a version of (237) that has no reading at all, It is given in (248).

(247a) Only WIDENER is opposite from Schoenhof's. (247b) *WIDENER is only opposite from Schoenhof's,

(248) *THE NEPTUNIANS could only have even read PLA YEOY.

Here, although even could be linked to PLA YEOY, only could not have been floated away from THE NEPTUNIANS, as *(247b) shows. Thus, *(248) is not starred on a band-crossing reading merely because of the existence of a stronger reading that would not entail band-crossing; (248) is ungrammatical on all readings.

Our conclusion, then, is not that (241) through (243) are better than (236) and (237) for transderivational reasons, but rather that quantifiers are less su bject than intensifiers to band-crossing constraints, which is due to the fact that quantifiers are more flexible than are intensifiers, as regards being separated by raising from a subject to which they are linked. This difference in flexibility we have noted already, in connection with examples (223) and (224).

The point of this section lies in the (less than perfect) parallelism between (236) and (237) and (241) through (243). The fact that both intensifiers and quantifiers can give rise to band-crossing violations is an indirect piece of evidence for our analysis of even and also, which we argue come to he nonadjacent to. and to the right of. focused subjects by the same processes that have this effect in the case of quantifiers.

Summary Remarks on Auxiliaries

Our conclusion is that the distribution of intensifiers with respect to aux iliary verbs is best accounted for if auxiliaries are analyzed as main verbs that trigger A-Raising. Our assessment of previous work on intensifiers suggests

of such sentences as (193) through (196) and (224c), we argued for a cyclic analysis of even and also, in which these words are moved rightward from focused subjects by an expanded rule of Quantifier Floating. The rules in question are stated in the following section.

Conclusions on Like and Intensifiers

We can summarize the discussion of like and its neighbors by observing that the similarities in behavior among like, even, also, only, just, and too far outweigh the differences. In fact, a very large amount of otherwise inexplicable data can be handled by the ordered application of the following general rules:

(249) I. Intensifier Hopping X Intensifier A Y I 2 3 4 OPT I 0 3+,+2+,4 ~

Conditions: (a) blocked if A is on a left branch of some other constituent than S.

(b) obligatory for too (c) blocks for just (d) A is a constituent that contains the focused

element (249) 2. Nuclear Stress Rule

Vowel - [I stress] I [##[1 stress] Y## Condition: Y oF [I stress]

(249) 3. Quantifier Floating (optional and cyclic): In subject position, a NP that is "modified by" all, both, each, even, or also can ascend to become the derived subject, with the former modifiers being attached to the verb.»

(249) 4. Intensifier Climb A. W - [X - [Intensifier - A] - Y] .. Z

B A A B J 2 3 4 5 6 OPT 1311[2 0 4 5] 6 ~

!lThis rule is stated in words. rather than as a transformation. because il is intended as an approximation to a relational ascension rule. For the formal apparatus for stating relation­changing laws. see Perlmutter and Postal (/978).

i » _

III > > ~ az2.~~ M

/'

412 h ......::iS AND COOPER

B. W - [ X - [ A - Intensifier] - Y ] .. Z B A A B 2 3 4 5 6 OPT

I [2 3 0 5]1t4 6 => Conditions: (a) obligatory for just and also in PPs

(b) obligatory for all intensifiers if 3 + 4 = N (c) Rule 4B cannot apply if term 4 = only

(249) 5. Intensifier In X - [ Intensifier [ NP - V Y ] ] - Z

S S S S 2 3 4 5

I 0 3+2 4 5 => Conditions: (a) obligatory for even and only

(b) blocks for like (c) preferred for also

The ordering relations that obtain amongthese rules are shown by the lines drawn in (250):

(250) (I.

e·3.

(4.5.

where a line between two rules means than the two rules must apply in the order given in (250), and where no line means that the rules can apply in either order. Rules I and 4 and 4 and 5 are intrinsically ordered, but it appears that Rules I and 2 must be extrinsically ordered.

We wish to extend the claim in (180) about the remote structure location of even to all intensifiers except like:

(251) The intensifiers even, also/too. and only/just can modify any semantic predicate or argument.

The reason for excluding like is that it appears to be able to modify in remote structure an even wider class of items than can the other intensifiers [d. (252)].

like ·cven

*(right) up."(252) Jcnuy picked it *also *only

.tt.\Vc note in passing that vernal particles such as lip, GU'Gr. out. in, along, etc, can nnly he modified by /ike if they arc preceded by right. We have 1\0 explanation for this puzzling [act.

11. LIKE SYNTAX 413

In addition, although we assume that tenses are to be represented in semantic structure as predicates and should thus be modifiable by all intensifiers, some are worse than others:

like only

(253) Helen { ?just ~ WAS a spy. ??also ??even

The general tendency seems to be for like to be able to occur in all environments where the other intensifiers can and in some others as well. In particular, it appears that like is freer than other intensifiers. in that it can be associated with any type of constituent, including NP, VP. N, V, Adj, Adv, Prep, Det, and Aux. Some of the more exotic of like's associations follow.

(254a) He's not only A bass player, he's like THE bass player. (254b) Jill's not ABOVE the car, she's like UNDER it. (254c) Tom gambled yesterday, and he like WILL gamble tomorrow.

Other intensifiers can be associated with most types of constituents, but not the full range exemplified by like, as shown below:

like

(255) Bob's *even "also

FATHER went to Chicago with us. (association with N)

*only

We noted at the outset of this paper that like can sometimes take the meaning of approximately, as in (256).

(256) I left at like nine 0 'clock.

Thus. approximately or about could replace like in (255) without a change in meaning. Approximate(1' and about arc more restricted than is like, however. in that they cannot modify a relational term, as in (257).

like }(257) I left *approximately early.

*about1 On the other hand. fillers having similar meaningsuch as kinda, sorta, and really require a relational term.

414 ROSS AND COOPER

kinda} (258a) I left t early.{ sor a

kinda} .(258b) *I left at { t rune 0 'clock.

sor a

It, thus, appears that like as a filler meaning approximately is less restricted than other words having the same meaning. This parallels the less restricted behavior of like as an-intensifier as compared with even, only, just, also, and

too.3'

LIKE AND MOVEMENT RULES

If like is associated with focus, as argued here, then its behavior should provide a litmus test for focus assignment in cases for which such assignment is debatable. In most of our examples to this point, we have relied on question-answer sentences where the focus is generally agreed upon. Now, however, we wish to extend the analysis of like to cases in which the location of the primary focus is less clear.

Langacker (1974), among others has claimed that fronting and backing rules serve complementary functions. Whereas fronting rules place focus on the moved constituent, backing rules reduce focus. Based on the prior discussion of like's association with focus, we predict that pauseless like should be allowed to precede moved constituents when fronted, but not when backed. This prediction is confirmed below in the case of left- vs. right­

dislocation.

(259a) Like John's stereo, it really bothers me.

(259b) *It really bothers me, like John's stereo.

Also consider topicalization (fronting) vs. heavy noun-phrase shift (backing).

(260a) Like these trivial points I'd never bring to Morgan's attention. (260b) 711 'd never bring to Morgan's attention like these trivial points.

A preference for like with fronted constituents is also observed for some adverbs, as shown below; but, for reasons that we do not understand, like can occur equally well with fronted vs, nonfronted prepositional phrases:

Fit seems particularly difficult to distinguish onlv andjusl semantically. yet we have observed striking differences in their behavior. If a relevant semantic difference cannot be found. we IlIUq

turn to phonology for an explanation of their different behavior. In this regard. note that hopped like and e"en require pauses. whereas hopped also. 011(1', and 100 do not. The members of the lailer group each terminate ill all open syllublc. unlike members of the lormer. We do not dare. however. suggest rhat thc pholloillgical distinction between open and closed syllables will lead to an account of why only some hopped intensifiers, in fact, require pauses.

11, LIKE SYNTAX 415

(261a) Like maybe we should leave tomorrow. (261 b) *We should leave tomorrow like rnaybe., (262a) Like in Chicago I'll attend the hardware convention. (262b) I'll attend tile hardware convention like in Chicago.

Using like as a means of ascribing focus has thus produced support (although not unanimous) for Langacker's claim about the functional distinction between fronting and backing rules.

SPEECH PROCESSING

At the outset, we hoped that the study of like, occurring as a filler in casual speech, would shed light on the nature of the speaker's code. Our findings suggest that this code contains an intricate syntactic component, similar to one required for intensifiers found in both speech and writing.

From the standpoint of speech production, the salient feature of this study is the systematic relation between like's position and the presence or absence of pausing. This finding, also generalized to intensifiers, raises the question: "Why is like accompanied by pausing when it follows the focused element but not when like precedes?" We have not been able to arrive at a satisfactory answer, although an account of this highly systematic relation must surely be included in any comprehensive theory of speech production. One possibility is that the pausing that accompanies hopped like and most intensifiers is produced as a consequence of the speaker's programming a relatively low fundamental frequency for these hopped words. Typically, the focused element carries a relatively high fundamental frequency (Fa), bearing stress. and the speaker might need to insert a short pause in order to make the laryngeal adjustments required for a subsequent low Fa(for a review of these laryngeal adjustments, see Halle& Stevens, 1971). In order for this account to hold, it must also explain why like is not accompanied by pauses in most circumstances when it precedes the focused element. To handle this fact. it would be necessary to motivate a principle stating that abrupt Falowering (of the kind that accompanies hopped like) requires more time to make laryngeal adjustments than the gradual Fa raising that accompanies pauseless like. No strong motivation is apparent, however, and, even if it were obtained. the account would leave unexplained why apause also occursjust after like when it follows the focused element.

The pauses that occur both before and after hopped like seem related to the pauses that accompany parentheticals. As with like. parenthetical expres­sions such as ",I guess" are bounded by pauses when such expressions follow the constituent they are associated with, as in (263b).

(263a) I guess HARRY flunked the writing exam.

416 ROSS AND COOPER

HARR Y, I guess, flunked the writing exam.

(263c) Harry is I guess FLUNKING the writing exam. Harry is FLUNKING, I guess, the writing exam.

(263b)

(263d)

In short, it seems like a principled account .of why pausing accompanies hopped like and intensifiers should also be capable of handling the pausing that accompanies "hopped" parentheticals (Cooper, in press; Emonds, 1976).

The present study also appears to have implications for speech perception, Insome circumstances, like may be inserted by speakers as an aid to listeners, signalling focus, Like's occurrence in such instances might produce a momentary heightening of attention in listeners, enabling them to process the focused element more efficiently (whether this element has just entered the perceptual system, as with hopped like, or is about to be entered, as with pauseless like), The listeners' heightened attention might be revealed experimentally during the performance of perceptual tasks such as phoneme­monitoring (F oss, 1969) or listening for mispronunciations (Cole, 1973). It is also possible that like's insertion, especially during very fast speech, serves as an aid to listeners by allowing them an extra fraction of a second to catch up in processing the input, as might be revealed in performance on tasks such as those utilized by Chodorow (Chapter 3, this volume).

. Finally, speakers themselves might be more likely to insert like when they are speaking to listeners who appear generally inattentive, With some stalwart like users, this possibility does not hold. For them, like is inserted during fast speech regardless of how the speaker perceives the listener's inattentiveness (in some cases, the speaker may not even be aware of the listener's state). But for those speakers who use like less frequently and who are sensitive to their listeners' attentiveness, like may be inserted as an attention-getting device. This possibility can be tested experimentally by examining conversations in which the listener's attentiveness is systematically

varied, In summary, this chapter has raised a number of questions about speech

processing, and certain of these questions have been sufficiently well defined' to permit subjecting them to empirical test. An ongoing interplay between armchair and laboratory seems to be the most promising way of continuing this study of casual speech. And so the moment has arrived for us to don our

laboratory whites.

REFERENCES

Anderson, S. R. flow to gel ('I'(·n. /.<lII/:II<1[:e. 1972,48. R93·906. Baker, C. L. Stress bel and auxiliary behavior in English. Linguistic Inquiry, 1972,2. 167 IRI. Boomer. D. S, Hesitation and grammatical encoding. Language and Speech, 1965.8. 148..15R. Bresnan, J. W. Sentence stress and syntactic transformations. Language. 197\, 47,257 ..281.

11. LIKE SYNTAX 411

Chodorow, M. S. Time compressed speech and the study of lexical and syntactic processing. In W. E. Cooper & E. C. T. Walker (Eds.), Sentence processing: Psycholinguistic studies presented to Merrill Garrell. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1979.

Chomsky, N. Deep structure, surface structure, and semantic interpretation. In D. Steinberg & L. Jakobovitz (Eds.), Semantics: An interdisciplinary reader in philosophy. linguistics, and psychology. New York: Cambridge University Press, 1971.

Chomsky, N.,·& Halle, M. The sound pattern of English. New York: Harper & Row, 19(,R. Cole, R. A. Listening for mispronunciations. Perception & Psychophysics. 1973, 14. 153-156. Cooper, W. E. Syntactic-to-phonetic coding. In B. Butterworth (Ed.), Language production.

New York: Academic Press, in press. Emends, J. E. A transformational approach to English syntax. New York: Academic Press,

1976. Fauconnier, G. Do quantifiers branch? Linguistic Inquiry. 1975,6. 555-567. Fiengo, R. W., & Lasnik, H. Some issues in the theory of transformations. Linguistic Inquiry.

1976,7. 182-191. Foss, D. J. Decision processes during sentence comprehension: Effects of lexical item difficulty

and position upon decision times. Journal of Verbal Learning and Verbal Behavior, 1969,8. 457-462.

Fraser, B. J. An analysis of 'even' in English. In C. Fillmore & D. T. Langendoen (Eds.), Studies in linguistic semantics. New York: Holt, Rinehart, & Winston, 1971.

Fromkin, V. A. The non-anomalous nature of anomalous utterances. Language, 1971,47. 27-52.

Garrett, M. F. The analysis of sentence production. In G. Bower (Ed.), Advances in learning theory and motivation (Vol. 9). New York: Academic Press, 1975.

Goldman-Eisler, F. Psycholinguistics: Experiments in spontaneous speech. New York: Academic Press, 1968.

Halle, M., & Stevens, K. N. A note on laryngeal features. Quarterly Progress Report of the M.I. T. Research Laboratory of Electronics. 1971, 101, 198-213.

Hankamer, J. Unacceptable ambiguity. Linguistic Inquiry, 1973, 4. 17-68. Horn, L. R. A presuppositional analysis of only and even. In R. I. Binnick, A. Davidson, G.

Green. & J. Morgan (Eds.), Papers/rom the fifth regional meeting ofthe Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 1969.

Jackendoff, R. S. Semantic interpretation in generative grammar. Cambridge, Mass.: MIT Press, 1972.

James, D. The syntax and semantics of interjections in English. Unpublished doctoral dissertation, University of Michigan, Ann Arbor, 1973.

Kayne, R. French syntax: The transformational cycle. Cambridge, Mass.: MIT Press, 1975. Keyser, S. J .. & Postal, P. M. Beginning Englishgrammar. New York: Harper & Row, 1976. King, .H. V. On blocking the rules for contraction in English. Linguistic Inquiry. 1970. I.

134-136. Klima, E. S. Negation in English. In J. A. Fodor & J. J. Katz (Eds.), Thestructure oflanguage:

Readings in the philosophy oflanguage. Englewood Cliffs, N.J.: Prentice-II all, 1964. Kuroda, S-Y. Generative grammatical studies in the Japanese language. Unpublished doctoral

dissertation, MIT, 1965. Langacker, R. W. Mirror image rules I: Syntax. Language, 1969,45, 575-598. Langacker, R. W. Movement rules in functional perspective. Language, 1974,50. 630-664. Maling, J. Notes on quantifier-postposing, Linguistic Inquiry. 1976,7. 708..718. Martin, J. G. Some acoustic and grammatical features of spontaneous speech. In D. J. Horton &

J. J. Jenkins (Eds.), The perception oflanguage. Columbus, Ohio: C. E. Merrill. 1971. Perlmutter, D. M., & Postal. P. M. Relational grammar. Book in preparation, 197R. Postal, P. M. On raising: One rule of English grammar and its theoretical implications.

Cambridge, Mass.: MIT Press, 1974. (a)

418 ROSS AND COOPER

Postal, P. M. On certain ambiguities. Linguistic Inquiry. 1974,5, 367-424. (b) Reinhart, T. The syntactic domain ofanaphora. Unpublisheddoctoral dissertation. MIT, 1976. Ross. J. R. Constraints on variables in syntax. Unpublisheddoctoral dissertation. MIT, 1967. Ross, J. R. Auxiliaries as main verbs. In W. Todd (Ed.), STudies in philosophical linguistics-«

Series one. Evanston, Ill.: Great Expectations, 1969. Ross, J. R. Clausematiness.ln E. L. Keenan(Ed.), Formal semantics ofnatural language. New

York: Cambridge University Press, 1975. Shattuck-Hufnagel, S. Remarks on an aspect of a sentenceproduction model based on speech­

error patterns. In W. E. Cooper & E. C. T. Walker (Eds.), Sentence processing: Psycholinguistic studies presented TO Merrill Garrell. Hillsdale, N.J.: Lawrence Erlbaurn Associates. 1979. •

Stampe, D. How I spent my summer vacation. Unpublisheddoctoral dissertation, Universityof Chicago. 1972.

Zwicky, A. M. Auxiliary reduction in English. Linguistic Inquiry, 1970,1.323-336.

Three Cheers for Propositional Attitudes (Some Reflections on D. C. Dennett's "Intentional Systems")

J. A. Fodor Massachusetts Institute of Technology

The belief that P has fal1en upon hard times.' Here is how it came about. Many respectable philosophers used to think that substitution instances of the scheme y believes tht P (=S), together with their friends and logical relations, are eliminable from English salva the expressive power of the language. Roughly; each sentence generated by substitution in S was to be replaced by some (logical1y equivalent, perhaps synonymous) sentence in which the predicate expresses a form of behavior or a disposition to behave. Insouciance prevailed; nobody actual1y provided the analyses, and there were those who seemed to glory in not providing them. So, Ryle (iff read him right) held both that talking about propositional attitudes is just a way of talking about behavior or behavioral dispositions and that nothing that can be (finitely) said about behavior or behavioral dispositions can, in the general case, exhaust the content of the ascription of a propositional attitude. A darkish doctrine, no doubt, but those were darkish times.

At their very worst, however. the facticity of such ascriptions went unimpugned. If, after al1, claims about the belief that P arc claims about the

II use'believe'as my paradigm of a verb of propositionalattitude, but sometimesvarythe verb to break the tedium. What I assume that all such verbs have in common(insofar as they engage the topic of this paper) is their tendency to establish opaque (specifically, intentional) contexts. Opaque contexts are. for example, those in whichthe substitution of corererringexpressiOIlS fails to preservetruth. That isa miserably imprecisecharacterization or the classor casesat issue, hut no more precision is needed for the purposes at hand.

A note on orthography: I USe 'intensional' to mean, in effect,opaque. and I usc'intenrional' In

mean opaque and psychological. Some intensionalcontexts are thus nonintentional in this usage: whereas. all intentional contexts are ipso facto intensional.

419

I I

(0',